?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'vim-with-runtime' into feat/tagfunc
[vim_extended.git] / src / regexp.c
blob53559455e0cfd372d295a6d14ccb64771b48d926
1 /* vi:set ts=8 sts=4 sw=4:
2 *
3 * Handling of regular expressions: vim_regcomp(), vim_regexec(), vim_regsub()
4 *
5 * NOTICE:
6 *
7 * This is NOT the original regular expression code as written by Henry
8 * Spencer. This code has been modified specifically for use with the VIM
9 * editor, and should not be used separately from Vim. If you want a good
10 * regular expression library, get the original code. The copyright notice
11 * that follows is from the original.
12 *
13 * END NOTICE
14 *
15 * Copyright (c) 1986 by University of Toronto.
16 * Written by Henry Spencer. Not derived from licensed software.
17 *
18 * Permission is granted to anyone to use this software for any
19 * purpose on any computer system, and to redistribute it freely,
20 * subject to the following restrictions:
21 *
22 * 1. The author is not responsible for the consequences of use of
23 * this software, no matter how awful, even if they arise
24 * from defects in it.
25 *
26 * 2. The origin of this software must not be misrepresented, either
27 * by explicit claim or by omission.
28 *
29 * 3. Altered versions must be plainly marked as such, and must not
30 * be misrepresented as being the original software.
31 *
32 * Beware that some of this code is subtly aware of the way operator
33 * precedence is structured in regular expressions. Serious changes in
34 * regular-expression syntax might require a total rethink.
35 *
36 * Changes have been made by Tony Andrews, Olaf 'Rhialto' Seibert, Robert
37 * Webb, Ciaran McCreesh and Bram Moolenaar.
38 * Named character class support added by Walter Briscoe (1998 Jul 01)
39 */
40
41 #include "vim.h"
42
43 #undef DEBUG
44
45 /*
46 * The "internal use only" fields in regexp.h are present to pass info from
47 * compile to execute that permits the execute phase to run lots faster on
48 * simple cases. They are:
49 *
50 * regstart char that must begin a match; NUL if none obvious; Can be a
51 * multi-byte character.
52 * reganch is the match anchored (at beginning-of-line only)?
53 * regmust string (pointer into program) that match must include, or NULL
54 * regmlen length of regmust string
55 * regflags RF_ values or'ed together
56 *
57 * Regstart and reganch permit very fast decisions on suitable starting points
58 * for a match, cutting down the work a lot. Regmust permits fast rejection
59 * of lines that cannot possibly match. The regmust tests are costly enough
60 * that vim_regcomp() supplies a regmust only if the r.e. contains something
61 * potentially expensive (at present, the only such thing detected is * or +
62 * at the start of the r.e., which can involve a lot of backup). Regmlen is
63 * supplied because the test in vim_regexec() needs it and vim_regcomp() is
64 * computing it anyway.
65 */
66
67 /*
68 * Structure for regexp "program". This is essentially a linear encoding
69 * of a nondeterministic finite-state machine (aka syntax charts or
70 * "railroad normal form" in parsing technology). Each node is an opcode
71 * plus a "next" pointer, possibly plus an operand. "Next" pointers of
72 * all nodes except BRANCH and BRACES_COMPLEX implement concatenation; a "next"
73 * pointer with a BRANCH on both ends of it is connecting two alternatives.
74 * (Here we have one of the subtle syntax dependencies: an individual BRANCH
75 * (as opposed to a collection of them) is never concatenated with anything
76 * because of operator precedence). The "next" pointer of a BRACES_COMPLEX
77 * node points to the node after the stuff to be repeated.
78 * The operand of some types of node is a literal string; for others, it is a
79 * node leading into a sub-FSM. In particular, the operand of a BRANCH node
80 * is the first node of the branch.
81 * (NB this is *not* a tree structure: the tail of the branch connects to the
82 * thing following the set of BRANCHes.)
83 *
84 * pattern is coded like:
85 *
86 * +-----------------+
87 * | V
88 * <aa>\|<bb> BRANCH <aa> BRANCH <bb> --> END
89 * | ^ | ^
90 * +------+ +----------+
91 *
92 *
93 * +------------------+
94 * V |
95 * <aa>* BRANCH BRANCH <aa> --> BACK BRANCH --> NOTHING --> END
96 * | | ^ ^
97 * | +---------------+ |
98 * +---------------------------------------------+
99 *
100 *
101 * +----------------------+
102 * V |
103 * <aa>\+ BRANCH <aa> --> BRANCH --> BACK BRANCH --> NOTHING --> END
104 * | | ^ ^
105 * | +-----------+ |
106 * +--------------------------------------------------+
107 *
108 *
109 * +-------------------------+
110 * V |
111 * <aa>\{} BRANCH BRACE_LIMITS --> BRACE_COMPLEX <aa> --> BACK END
112 * | | ^
113 * | +----------------+
114 * +-----------------------------------------------+
115 *
116 *
117 * <aa>\@!<bb> BRANCH NOMATCH <aa> --> END <bb> --> END
118 * | | ^ ^
119 * | +----------------+ |
120 * +--------------------------------+
121 *
122 * +---------+
123 * | V
124 * \z[abc] BRANCH BRANCH a BRANCH b BRANCH c BRANCH NOTHING --> END
125 * | | | | ^ ^
126 * | | | +-----+ |
127 * | | +----------------+ |
128 * | +---------------------------+ |
129 * +------------------------------------------------------+
130 *
131 * They all start with a BRANCH for "\|" alternatives, even when there is only
132 * one alternative.
133 */
134
135 /*
136 * The opcodes are:
137 */
138
139 /* definition number opnd? meaning */
140 #define END 0 /* End of program or NOMATCH operand. */
141 #define BOL 1 /* Match "" at beginning of line. */
142 #define EOL 2 /* Match "" at end of line. */
143 #define BRANCH 3 /* node Match this alternative, or the
144 * next... */
145 #define BACK 4 /* Match "", "next" ptr points backward. */
146 #define EXACTLY 5 /* str Match this string. */
147 #define NOTHING 6 /* Match empty string. */
148 #define STAR 7 /* node Match this (simple) thing 0 or more
149 * times. */
150 #define PLUS 8 /* node Match this (simple) thing 1 or more
151 * times. */
152 #define MATCH 9 /* node match the operand zero-width */
153 #define NOMATCH 10 /* node check for no match with operand */
154 #define BEHIND 11 /* node look behind for a match with operand */
155 #define NOBEHIND 12 /* node look behind for no match with operand */
156 #define SUBPAT 13 /* node match the operand here */
157 #define BRACE_SIMPLE 14 /* node Match this (simple) thing between m and
158 * n times (\{m,n\}). */
159 #define BOW 15 /* Match "" after [^a-zA-Z0-9_] */
160 #define EOW 16 /* Match "" at [^a-zA-Z0-9_] */
161 #define BRACE_LIMITS 17 /* nr nr define the min & max for BRACE_SIMPLE
162 * and BRACE_COMPLEX. */
163 #define NEWL 18 /* Match line-break */
164 #define BHPOS 19 /* End position for BEHIND or NOBEHIND */
165
166
167 /* character classes: 20-48 normal, 50-78 include a line-break */
168 #define ADD_NL 30
169 #define FIRST_NL ANY + ADD_NL
170 #define ANY 20 /* Match any one character. */
171 #define ANYOF 21 /* str Match any character in this string. */
172 #define ANYBUT 22 /* str Match any character not in this
173 * string. */
174 #define IDENT 23 /* Match identifier char */
175 #define SIDENT 24 /* Match identifier char but no digit */
176 #define KWORD 25 /* Match keyword char */
177 #define SKWORD 26 /* Match word char but no digit */
178 #define FNAME 27 /* Match file name char */
179 #define SFNAME 28 /* Match file name char but no digit */
180 #define PRINT 29 /* Match printable char */
181 #define SPRINT 30 /* Match printable char but no digit */
182 #define WHITE 31 /* Match whitespace char */
183 #define NWHITE 32 /* Match non-whitespace char */
184 #define DIGIT 33 /* Match digit char */
185 #define NDIGIT 34 /* Match non-digit char */
186 #define HEX 35 /* Match hex char */
187 #define NHEX 36 /* Match non-hex char */
188 #define OCTAL 37 /* Match octal char */
189 #define NOCTAL 38 /* Match non-octal char */
190 #define WORD 39 /* Match word char */
191 #define NWORD 40 /* Match non-word char */
192 #define HEAD 41 /* Match head char */
193 #define NHEAD 42 /* Match non-head char */
194 #define ALPHA 43 /* Match alpha char */
195 #define NALPHA 44 /* Match non-alpha char */
196 #define LOWER 45 /* Match lowercase char */
197 #define NLOWER 46 /* Match non-lowercase char */
198 #define UPPER 47 /* Match uppercase char */
199 #define NUPPER 48 /* Match non-uppercase char */
200 #define LAST_NL NUPPER + ADD_NL
201 #define WITH_NL(op) ((op) >= FIRST_NL && (op) <= LAST_NL)
202
203 #define MOPEN 80 /* -89 Mark this point in input as start of
204 * \( subexpr. MOPEN + 0 marks start of
205 * match. */
206 #define MCLOSE 90 /* -99 Analogous to MOPEN. MCLOSE + 0 marks
207 * end of match. */
208 #define BACKREF 100 /* -109 node Match same string again \1-\9 */
209
210 #ifdef FEAT_SYN_HL
211 # define ZOPEN 110 /* -119 Mark this point in input as start of
212 * \z( subexpr. */
213 # define ZCLOSE 120 /* -129 Analogous to ZOPEN. */
214 # define ZREF 130 /* -139 node Match external submatch \z1-\z9 */
215 #endif
216
217 #define BRACE_COMPLEX 140 /* -149 node Match nodes between m & n times */
218
219 #define NOPEN 150 /* Mark this point in input as start of
220 \%( subexpr. */
221 #define NCLOSE 151 /* Analogous to NOPEN. */
222
223 #define MULTIBYTECODE 200 /* mbc Match one multi-byte character */
224 #define RE_BOF 201 /* Match "" at beginning of file. */
225 #define RE_EOF 202 /* Match "" at end of file. */
226 #define CURSOR 203 /* Match location of cursor. */
227
228 #define RE_LNUM 204 /* nr cmp Match line number */
229 #define RE_COL 205 /* nr cmp Match column number */
230 #define RE_VCOL 206 /* nr cmp Match virtual column number */
231
232 #define RE_MARK 207 /* mark cmp Match mark position */
233 #define RE_VISUAL 208 /* Match Visual area */
234
235 /*
236 * Magic characters have a special meaning, they don't match literally.
237 * Magic characters are negative. This separates them from literal characters
238 * (possibly multi-byte). Only ASCII characters can be Magic.
239 */
240 #define Magic(x) ((int)(x) - 256)
241 #define un_Magic(x) ((x) + 256)
242 #define is_Magic(x) ((x) < 0)
243
244 static int no_Magic __ARGS((int x));
245 static int toggle_Magic __ARGS((int x));
246
247 static int
248 no_Magic(x)
249 int x;
250 {
251 if (is_Magic(x))
252 return un_Magic(x);
253 return x;
254 }
255
256 static int
257 toggle_Magic(x)
258 int x;
259 {
260 if (is_Magic(x))
261 return un_Magic(x);
262 return Magic(x);
263 }
264
265 /*
266 * The first byte of the regexp internal "program" is actually this magic
267 * number; the start node begins in the second byte. It's used to catch the
268 * most severe mutilation of the program by the caller.
269 */
270
271 #define REGMAGIC 0234
272
273 /*
274 * Opcode notes:
275 *
276 * BRANCH The set of branches constituting a single choice are hooked
277 * together with their "next" pointers, since precedence prevents
278 * anything being concatenated to any individual branch. The
279 * "next" pointer of the last BRANCH in a choice points to the
280 * thing following the whole choice. This is also where the
281 * final "next" pointer of each individual branch points; each
282 * branch starts with the operand node of a BRANCH node.
283 *
284 * BACK Normal "next" pointers all implicitly point forward; BACK
285 * exists to make loop structures possible.
286 *
287 * STAR,PLUS '=', and complex '*' and '+', are implemented as circular
288 * BRANCH structures using BACK. Simple cases (one character
289 * per match) are implemented with STAR and PLUS for speed
290 * and to minimize recursive plunges.
291 *
292 * BRACE_LIMITS This is always followed by a BRACE_SIMPLE or BRACE_COMPLEX
293 * node, and defines the min and max limits to be used for that
294 * node.
295 *
296 * MOPEN,MCLOSE ...are numbered at compile time.
297 * ZOPEN,ZCLOSE ...ditto
298 */
299
300 /*
301 * A node is one char of opcode followed by two chars of "next" pointer.
302 * "Next" pointers are stored as two 8-bit bytes, high order first. The
303 * value is a positive offset from the opcode of the node containing it.
304 * An operand, if any, simply follows the node. (Note that much of the
305 * code generation knows about this implicit relationship.)
306 *
307 * Using two bytes for the "next" pointer is vast overkill for most things,
308 * but allows patterns to get big without disasters.
309 */
310 #define OP(p) ((int)*(p))
311 #define NEXT(p) (((*((p) + 1) & 0377) << 8) + (*((p) + 2) & 0377))
312 #define OPERAND(p) ((p) + 3)
313 /* Obtain an operand that was stored as four bytes, MSB first. */
314 #define OPERAND_MIN(p) (((long)(p)[3] << 24) + ((long)(p)[4] << 16) \
315 + ((long)(p)[5] << 8) + (long)(p)[6])
316 /* Obtain a second operand stored as four bytes. */
317 #define OPERAND_MAX(p) OPERAND_MIN((p) + 4)
318 /* Obtain a second single-byte operand stored after a four bytes operand. */
319 #define OPERAND_CMP(p) (p)[7]
320
321 /*
322 * Utility definitions.
323 */
324 #define UCHARAT(p) ((int)*(char_u *)(p))
325
326 /* Used for an error (down from) vim_regcomp(): give the error message, set
327 * rc_did_emsg and return NULL */
328 #define EMSG_RET_NULL(m) return (EMSG(m), rc_did_emsg = TRUE, (void *)NULL)
329 #define EMSG_M_RET_NULL(m, c) return (EMSG2((m), (c) ? "" : "\\"), rc_did_emsg = TRUE, (void *)NULL)
330 #define EMSG_RET_FAIL(m) return (EMSG(m), rc_did_emsg = TRUE, FAIL)
331 #define EMSG_ONE_RET_NULL EMSG_M_RET_NULL(_("E369: invalid item in %s%%[]"), reg_magic == MAGIC_ALL)
332
333 #define MAX_LIMIT (32767L << 16L)
334
335 static int re_multi_type __ARGS((int));
336 static int cstrncmp __ARGS((char_u *s1, char_u *s2, int *n));
337 static char_u *cstrchr __ARGS((char_u *, int));
338
339 #ifdef DEBUG
340 static void regdump __ARGS((char_u *, regprog_T *));
341 static char_u *regprop __ARGS((char_u *));
342 #endif
343
344 #define NOT_MULTI 0
345 #define MULTI_ONE 1
346 #define MULTI_MULT 2
347 /*
348 * Return NOT_MULTI if c is not a "multi" operator.
349 * Return MULTI_ONE if c is a single "multi" operator.
350 * Return MULTI_MULT if c is a multi "multi" operator.
351 */
352 static int
353 re_multi_type(c)
354 int c;
355 {
356 if (c == Magic('@') || c == Magic('=') || c == Magic('?'))
357 return MULTI_ONE;
358 if (c == Magic('*') || c == Magic('+') || c == Magic('{'))
359 return MULTI_MULT;
360 return NOT_MULTI;
361 }
362
363 /*
364 * Flags to be passed up and down.
365 */
366 #define HASWIDTH 0x1 /* Known never to match null string. */
367 #define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
368 #define SPSTART 0x4 /* Starts with * or +. */
369 #define HASNL 0x8 /* Contains some \n. */
370 #define HASLOOKBH 0x10 /* Contains "\@<=" or "\@<!". */
371 #define WORST 0 /* Worst case. */
372
373 /*
374 * When regcode is set to this value, code is not emitted and size is computed
375 * instead.
376 */
377 #define JUST_CALC_SIZE ((char_u *) -1)
378
379 static char_u *reg_prev_sub = NULL;
380
381 /*
382 * REGEXP_INRANGE contains all characters which are always special in a []
383 * range after '\'.
384 * REGEXP_ABBR contains all characters which act as abbreviations after '\'.
385 * These are:
386 * \n - New line (NL).
387 * \r - Carriage Return (CR).
388 * \t - Tab (TAB).
389 * \e - Escape (ESC).
390 * \b - Backspace (Ctrl_H).
391 * \d - Character code in decimal, eg \d123
392 * \o - Character code in octal, eg \o80
393 * \x - Character code in hex, eg \x4a
394 * \u - Multibyte character code, eg \u20ac
395 * \U - Long multibyte character code, eg \U12345678
396 */
397 static char_u REGEXP_INRANGE[] = "]^-n\\";
398 static char_u REGEXP_ABBR[] = "nrtebdoxuU";
399
400 static int backslash_trans __ARGS((int c));
401 static int get_char_class __ARGS((char_u **pp));
402 static int get_equi_class __ARGS((char_u **pp));
403 static void reg_equi_class __ARGS((int c));
404 static int get_coll_element __ARGS((char_u **pp));
405 static char_u *skip_anyof __ARGS((char_u *p));
406 static void init_class_tab __ARGS((void));
407
408 /*
409 * Translate '\x' to its control character, except "\n", which is Magic.
410 */
411 static int
412 backslash_trans(c)
413 int c;
414 {
415 switch (c)
416 {
417 case 'r': return CAR;
418 case 't': return TAB;
419 case 'e': return ESC;
420 case 'b': return BS;
421 }
422 return c;
423 }
424
425 /*
426 * Check for a character class name "[:name:]". "pp" points to the '['.
427 * Returns one of the CLASS_ items. CLASS_NONE means that no item was
428 * recognized. Otherwise "pp" is advanced to after the item.
429 */
430 static int
431 get_char_class(pp)
432 char_u **pp;
433 {
434 static const char *(class_names[]) =
435 {
436 "alnum:]",
437 #define CLASS_ALNUM 0
438 "alpha:]",
439 #define CLASS_ALPHA 1
440 "blank:]",
441 #define CLASS_BLANK 2
442 "cntrl:]",
443 #define CLASS_CNTRL 3
444 "digit:]",
445 #define CLASS_DIGIT 4
446 "graph:]",
447 #define CLASS_GRAPH 5
448 "lower:]",
449 #define CLASS_LOWER 6
450 "print:]",
451 #define CLASS_PRINT 7
452 "punct:]",
453 #define CLASS_PUNCT 8
454 "space:]",
455 #define CLASS_SPACE 9
456 "upper:]",
457 #define CLASS_UPPER 10
458 "xdigit:]",
459 #define CLASS_XDIGIT 11
460 "tab:]",
461 #define CLASS_TAB 12
462 "return:]",
463 #define CLASS_RETURN 13
464 "backspace:]",
465 #define CLASS_BACKSPACE 14
466 "escape:]",
467 #define CLASS_ESCAPE 15
468 };
469 #define CLASS_NONE 99
470 int i;
471
472 if ((*pp)[1] == ':')
473 {
474 for (i = 0; i < (int)(sizeof(class_names) / sizeof(*class_names)); ++i)
475 if (STRNCMP(*pp + 2, class_names[i], STRLEN(class_names[i])) == 0)
476 {
477 *pp += STRLEN(class_names[i]) + 2;
478 return i;
479 }
480 }
481 return CLASS_NONE;
482 }
483
484 /*
485 * Specific version of character class functions.
486 * Using a table to keep this fast.
487 */
488 static short class_tab[256];
489
490 #define RI_DIGIT 0x01
491 #define RI_HEX 0x02
492 #define RI_OCTAL 0x04
493 #define RI_WORD 0x08
494 #define RI_HEAD 0x10
495 #define RI_ALPHA 0x20
496 #define RI_LOWER 0x40
497 #define RI_UPPER 0x80
498 #define RI_WHITE 0x100
499
500 static void
501 init_class_tab()
502 {
503 int i;
504 static int done = FALSE;
505
506 if (done)
507 return;
508
509 for (i = 0; i < 256; ++i)
510 {
511 if (i >= '0' && i <= '7')
512 class_tab[i] = RI_DIGIT + RI_HEX + RI_OCTAL + RI_WORD;
513 else if (i >= '8' && i <= '9')
514 class_tab[i] = RI_DIGIT + RI_HEX + RI_WORD;
515 else if (i >= 'a' && i <= 'f')
516 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
517 #ifdef EBCDIC
518 else if ((i >= 'g' && i <= 'i') || (i >= 'j' && i <= 'r')
519 || (i >= 's' && i <= 'z'))
520 #else
521 else if (i >= 'g' && i <= 'z')
522 #endif
523 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_LOWER;
524 else if (i >= 'A' && i <= 'F')
525 class_tab[i] = RI_HEX + RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
526 #ifdef EBCDIC
527 else if ((i >= 'G' && i <= 'I') || ( i >= 'J' && i <= 'R')
528 || (i >= 'S' && i <= 'Z'))
529 #else
530 else if (i >= 'G' && i <= 'Z')
531 #endif
532 class_tab[i] = RI_WORD + RI_HEAD + RI_ALPHA + RI_UPPER;
533 else if (i == '_')
534 class_tab[i] = RI_WORD + RI_HEAD;
535 else
536 class_tab[i] = 0;
537 }
538 class_tab[' '] |= RI_WHITE;
539 class_tab['\t'] |= RI_WHITE;
540 done = TRUE;
541 }
542
543 #ifdef FEAT_MBYTE
544 # define ri_digit(c) (c < 0x100 && (class_tab[c] & RI_DIGIT))
545 # define ri_hex(c) (c < 0x100 && (class_tab[c] & RI_HEX))
546 # define ri_octal(c) (c < 0x100 && (class_tab[c] & RI_OCTAL))
547 # define ri_word(c) (c < 0x100 && (class_tab[c] & RI_WORD))
548 # define ri_head(c) (c < 0x100 && (class_tab[c] & RI_HEAD))
549 # define ri_alpha(c) (c < 0x100 && (class_tab[c] & RI_ALPHA))
550 # define ri_lower(c) (c < 0x100 && (class_tab[c] & RI_LOWER))
551 # define ri_upper(c) (c < 0x100 && (class_tab[c] & RI_UPPER))
552 # define ri_white(c) (c < 0x100 && (class_tab[c] & RI_WHITE))
553 #else
554 # define ri_digit(c) (class_tab[c] & RI_DIGIT)
555 # define ri_hex(c) (class_tab[c] & RI_HEX)
556 # define ri_octal(c) (class_tab[c] & RI_OCTAL)
557 # define ri_word(c) (class_tab[c] & RI_WORD)
558 # define ri_head(c) (class_tab[c] & RI_HEAD)
559 # define ri_alpha(c) (class_tab[c] & RI_ALPHA)
560 # define ri_lower(c) (class_tab[c] & RI_LOWER)
561 # define ri_upper(c) (class_tab[c] & RI_UPPER)
562 # define ri_white(c) (class_tab[c] & RI_WHITE)
563 #endif
564
565 /* flags for regflags */
566 #define RF_ICASE 1 /* ignore case */
567 #define RF_NOICASE 2 /* don't ignore case */
568 #define RF_HASNL 4 /* can match a NL */
569 #define RF_ICOMBINE 8 /* ignore combining characters */
570 #define RF_LOOKBH 16 /* uses "\@<=" or "\@<!" */
571
572 /*
573 * Global work variables for vim_regcomp().
574 */
575
576 static char_u *regparse; /* Input-scan pointer. */
577 static int prevchr_len; /* byte length of previous char */
578 static int num_complex_braces; /* Complex \{...} count */
579 static int regnpar; /* () count. */
580 #ifdef FEAT_SYN_HL
581 static int regnzpar; /* \z() count. */
582 static int re_has_z; /* \z item detected */
583 #endif
584 static char_u *regcode; /* Code-emit pointer, or JUST_CALC_SIZE */
585 static long regsize; /* Code size. */
586 static int reg_toolong; /* TRUE when offset out of range */
587 static char_u had_endbrace[NSUBEXP]; /* flags, TRUE if end of () found */
588 static unsigned regflags; /* RF_ flags for prog */
589 static long brace_min[10]; /* Minimums for complex brace repeats */
590 static long brace_max[10]; /* Maximums for complex brace repeats */
591 static int brace_count[10]; /* Current counts for complex brace repeats */
592 #if defined(FEAT_SYN_HL) || defined(PROTO)
593 static int had_eol; /* TRUE when EOL found by vim_regcomp() */
594 #endif
595 static int one_exactly = FALSE; /* only do one char for EXACTLY */
596
597 static int reg_magic; /* magicness of the pattern: */
598 #define MAGIC_NONE 1 /* "\V" very unmagic */
599 #define MAGIC_OFF 2 /* "\M" or 'magic' off */
600 #define MAGIC_ON 3 /* "\m" or 'magic' */
601 #define MAGIC_ALL 4 /* "\v" very magic */
602
603 static int reg_string; /* matching with a string instead of a buffer
604 line */
605 static int reg_strict; /* "[abc" is illegal */
606
607 /*
608 * META contains all characters that may be magic, except '^' and '$'.
609 */
610
611 #ifdef EBCDIC
612 static char_u META[] = "%&()*+.123456789<=>?@ACDFHIKLMOPSUVWX[_acdfhiklmnopsuvwxz{|~";
613 #else
614 /* META[] is used often enough to justify turning it into a table. */
615 static char_u META_flags[] = {
616 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
617 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
618 /* % & ( ) * + . */
619 0, 0, 0, 0, 0, 1, 1, 0, 1, 1, 1, 1, 0, 0, 1, 0,
620 /* 1 2 3 4 5 6 7 8 9 < = > ? */
621 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1,
622 /* @ A C D F H I K L M O */
623 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1,
624 /* P S U V W X Z [ _ */
625 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 1,
626 /* a c d f h i k l m n o */
627 0, 1, 0, 1, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1,
628 /* p s u v w x z { | ~ */
629 1, 0, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1
630 };
631 #endif
632
633 static int curchr;
634
635 /* arguments for reg() */
636 #define REG_NOPAREN 0 /* toplevel reg() */
637 #define REG_PAREN 1 /* \(\) */
638 #define REG_ZPAREN 2 /* \z(\) */
639 #define REG_NPAREN 3 /* \%(\) */
640
641 /*
642 * Forward declarations for vim_regcomp()'s friends.
643 */
644 static void initchr __ARGS((char_u *));
645 static int getchr __ARGS((void));
646 static void skipchr_keepstart __ARGS((void));
647 static int peekchr __ARGS((void));
648 static void skipchr __ARGS((void));
649 static void ungetchr __ARGS((void));
650 static int gethexchrs __ARGS((int maxinputlen));
651 static int getoctchrs __ARGS((void));
652 static int getdecchrs __ARGS((void));
653 static int coll_get_char __ARGS((void));
654 static void regcomp_start __ARGS((char_u *expr, int flags));
655 static char_u *reg __ARGS((int, int *));
656 static char_u *regbranch __ARGS((int *flagp));
657 static char_u *regconcat __ARGS((int *flagp));
658 static char_u *regpiece __ARGS((int *));
659 static char_u *regatom __ARGS((int *));
660 static char_u *regnode __ARGS((int));
661 #ifdef FEAT_MBYTE
662 static int use_multibytecode __ARGS((int c));
663 #endif
664 static int prog_magic_wrong __ARGS((void));
665 static char_u *regnext __ARGS((char_u *));
666 static void regc __ARGS((int b));
667 #ifdef FEAT_MBYTE
668 static void regmbc __ARGS((int c));
669 #else
670 # define regmbc(c) regc(c)
671 #endif
672 static void reginsert __ARGS((int, char_u *));
673 static void reginsert_limits __ARGS((int, long, long, char_u *));
674 static char_u *re_put_long __ARGS((char_u *pr, long_u val));
675 static int read_limits __ARGS((long *, long *));
676 static void regtail __ARGS((char_u *, char_u *));
677 static void regoptail __ARGS((char_u *, char_u *));
678
679 /*
680 * Return TRUE if compiled regular expression "prog" can match a line break.
681 */
682 int
683 re_multiline(prog)
684 regprog_T *prog;
685 {
686 return (prog->regflags & RF_HASNL);
687 }
688
689 /*
690 * Return TRUE if compiled regular expression "prog" looks before the start
691 * position (pattern contains "\@<=" or "\@<!").
692 */
693 int
694 re_lookbehind(prog)
695 regprog_T *prog;
696 {
697 return (prog->regflags & RF_LOOKBH);
698 }
699
700 /*
701 * Check for an equivalence class name "[=a=]". "pp" points to the '['.
702 * Returns a character representing the class. Zero means that no item was
703 * recognized. Otherwise "pp" is advanced to after the item.
704 */
705 static int
706 get_equi_class(pp)
707 char_u **pp;
708 {
709 int c;
710 int l = 1;
711 char_u *p = *pp;
712
713 if (p[1] == '=')
714 {
715 #ifdef FEAT_MBYTE
716 if (has_mbyte)
717 l = (*mb_ptr2len)(p + 2);
718 #endif
719 if (p[l + 2] == '=' && p[l + 3] == ']')
720 {
721 #ifdef FEAT_MBYTE
722 if (has_mbyte)
723 c = mb_ptr2char(p + 2);
724 else
725 #endif
726 c = p[2];
727 *pp += l + 4;
728 return c;
729 }
730 }
731 return 0;
732 }
733
734 /*
735 * Produce the bytes for equivalence class "c".
736 * Currently only handles latin1, latin9 and utf-8.
737 */
738 static void
739 reg_equi_class(c)
740 int c;
741 {
742 #ifdef FEAT_MBYTE
743 if (enc_utf8 || STRCMP(p_enc, "latin1") == 0
744 || STRCMP(p_enc, "iso-8859-15") == 0)
745 #endif
746 {
747 switch (c)
748 {
749 case 'A': case '\300': case '\301': case '\302':
750 case '\303': case '\304': case '\305':
751 regmbc('A'); regmbc('\300'); regmbc('\301');
752 regmbc('\302'); regmbc('\303'); regmbc('\304');
753 regmbc('\305');
754 return;
755 case 'C': case '\307':
756 regmbc('C'); regmbc('\307');
757 return;
758 case 'E': case '\310': case '\311': case '\312': case '\313':
759 regmbc('E'); regmbc('\310'); regmbc('\311');
760 regmbc('\312'); regmbc('\313');
761 return;
762 case 'I': case '\314': case '\315': case '\316': case '\317':
763 regmbc('I'); regmbc('\314'); regmbc('\315');
764 regmbc('\316'); regmbc('\317');
765 return;
766 case 'N': case '\321':
767 regmbc('N'); regmbc('\321');
768 return;
769 case 'O': case '\322': case '\323': case '\324': case '\325':
770 case '\326':
771 regmbc('O'); regmbc('\322'); regmbc('\323');
772 regmbc('\324'); regmbc('\325'); regmbc('\326');
773 return;
774 case 'U': case '\331': case '\332': case '\333': case '\334':
775 regmbc('U'); regmbc('\331'); regmbc('\332');
776 regmbc('\333'); regmbc('\334');
777 return;
778 case 'Y': case '\335':
779 regmbc('Y'); regmbc('\335');
780 return;
781 case 'a': case '\340': case '\341': case '\342':
782 case '\343': case '\344': case '\345':
783 regmbc('a'); regmbc('\340'); regmbc('\341');
784 regmbc('\342'); regmbc('\343'); regmbc('\344');
785 regmbc('\345');
786 return;
787 case 'c': case '\347':
788 regmbc('c'); regmbc('\347');
789 return;
790 case 'e': case '\350': case '\351': case '\352': case '\353':
791 regmbc('e'); regmbc('\350'); regmbc('\351');
792 regmbc('\352'); regmbc('\353');
793 return;
794 case 'i': case '\354': case '\355': case '\356': case '\357':
795 regmbc('i'); regmbc('\354'); regmbc('\355');
796 regmbc('\356'); regmbc('\357');
797 return;
798 case 'n': case '\361':
799 regmbc('n'); regmbc('\361');
800 return;
801 case 'o': case '\362': case '\363': case '\364': case '\365':
802 case '\366':
803 regmbc('o'); regmbc('\362'); regmbc('\363');
804 regmbc('\364'); regmbc('\365'); regmbc('\366');
805 return;
806 case 'u': case '\371': case '\372': case '\373': case '\374':
807 regmbc('u'); regmbc('\371'); regmbc('\372');
808 regmbc('\373'); regmbc('\374');
809 return;
810 case 'y': case '\375': case '\377':
811 regmbc('y'); regmbc('\375'); regmbc('\377');
812 return;
813 }
814 }
815 regmbc(c);
816 }
817
818 /*
819 * Check for a collating element "[.a.]". "pp" points to the '['.
820 * Returns a character. Zero means that no item was recognized. Otherwise
821 * "pp" is advanced to after the item.
822 * Currently only single characters are recognized!
823 */
824 static int
825 get_coll_element(pp)
826 char_u **pp;
827 {
828 int c;
829 int l = 1;
830 char_u *p = *pp;
831
832 if (p[1] == '.')
833 {
834 #ifdef FEAT_MBYTE
835 if (has_mbyte)
836 l = (*mb_ptr2len)(p + 2);
837 #endif
838 if (p[l + 2] == '.' && p[l + 3] == ']')
839 {
840 #ifdef FEAT_MBYTE
841 if (has_mbyte)
842 c = mb_ptr2char(p + 2);
843 else
844 #endif
845 c = p[2];
846 *pp += l + 4;
847 return c;
848 }
849 }
850 return 0;
851 }
852
853
854 /*
855 * Skip over a "[]" range.
856 * "p" must point to the character after the '['.
857 * The returned pointer is on the matching ']', or the terminating NUL.
858 */
859 static char_u *
860 skip_anyof(p)
861 char_u *p;
862 {
863 int cpo_lit; /* 'cpoptions' contains 'l' flag */
864 int cpo_bsl; /* 'cpoptions' contains '\' flag */
865 #ifdef FEAT_MBYTE
866 int l;
867 #endif
868
869 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL;
870 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL;
871
872 if (*p == '^') /* Complement of range. */
873 ++p;
874 if (*p == ']' || *p == '-')
875 ++p;
876 while (*p != NUL && *p != ']')
877 {
878 #ifdef FEAT_MBYTE
879 if (has_mbyte && (l = (*mb_ptr2len)(p)) > 1)
880 p += l;
881 else
882 #endif
883 if (*p == '-')
884 {
885 ++p;
886 if (*p != ']' && *p != NUL)
887 mb_ptr_adv(p);
888 }
889 else if (*p == '\\'
890 && !cpo_bsl
891 && (vim_strchr(REGEXP_INRANGE, p[1]) != NULL
892 || (!cpo_lit && vim_strchr(REGEXP_ABBR, p[1]) != NULL)))
893 p += 2;
894 else if (*p == '[')
895 {
896 if (get_char_class(&p) == CLASS_NONE
897 && get_equi_class(&p) == 0
898 && get_coll_element(&p) == 0)
899 ++p; /* It was not a class name */
900 }
901 else
902 ++p;
903 }
904
905 return p;
906 }
907
908 /*
909 * Skip past regular expression.
910 * Stop at end of "startp" or where "dirc" is found ('/', '?', etc).
911 * Take care of characters with a backslash in front of it.
912 * Skip strings inside [ and ].
913 * When "newp" is not NULL and "dirc" is '?', make an allocated copy of the
914 * expression and change "\?" to "?". If "*newp" is not NULL the expression
915 * is changed in-place.
916 */
917 char_u *
918 skip_regexp(startp, dirc, magic, newp)
919 char_u *startp;
920 int dirc;
921 int magic;
922 char_u **newp;
923 {
924 int mymagic;
925 char_u *p = startp;
926
927 if (magic)
928 mymagic = MAGIC_ON;
929 else
930 mymagic = MAGIC_OFF;
931
932 for (; p[0] != NUL; mb_ptr_adv(p))
933 {
934 if (p[0] == dirc) /* found end of regexp */
935 break;
936 if ((p[0] == '[' && mymagic >= MAGIC_ON)
937 || (p[0] == '\\' && p[1] == '[' && mymagic <= MAGIC_OFF))
938 {
939 p = skip_anyof(p + 1);
940 if (p[0] == NUL)
941 break;
942 }
943 else if (p[0] == '\\' && p[1] != NUL)
944 {
945 if (dirc == '?' && newp != NULL && p[1] == '?')
946 {
947 /* change "\?" to "?", make a copy first. */
948 if (*newp == NULL)
949 {
950 *newp = vim_strsave(startp);
951 if (*newp != NULL)
952 p = *newp + (p - startp);
953 }
954 if (*newp != NULL)
955 STRMOVE(p, p + 1);
956 else
957 ++p;
958 }
959 else
960 ++p; /* skip next character */
961 if (*p == 'v')
962 mymagic = MAGIC_ALL;
963 else if (*p == 'V')
964 mymagic = MAGIC_NONE;
965 }
966 }
967 return p;
968 }
969
970 /*
971 * vim_regcomp() - compile a regular expression into internal code
972 * Returns the program in allocated space. Returns NULL for an error.
973 *
974 * We can't allocate space until we know how big the compiled form will be,
975 * but we can't compile it (and thus know how big it is) until we've got a
976 * place to put the code. So we cheat: we compile it twice, once with code
977 * generation turned off and size counting turned on, and once "for real".
978 * This also means that we don't allocate space until we are sure that the
979 * thing really will compile successfully, and we never have to move the
980 * code and thus invalidate pointers into it. (Note that it has to be in
981 * one piece because vim_free() must be able to free it all.)
982 *
983 * Whether upper/lower case is to be ignored is decided when executing the
984 * program, it does not matter here.
985 *
986 * Beware that the optimization-preparation code in here knows about some
987 * of the structure of the compiled regexp.
988 * "re_flags": RE_MAGIC and/or RE_STRING.
989 */
990 regprog_T *
991 vim_regcomp(expr, re_flags)
992 char_u *expr;
993 int re_flags;
994 {
995 regprog_T *r;
996 char_u *scan;
997 char_u *longest;
998 int len;
999 int flags;
1000
1001 if (expr == NULL)
1002 EMSG_RET_NULL(_(e_null));
1003
1004 init_class_tab();
1005
1006 /*
1007 * First pass: determine size, legality.
1008 */
1009 regcomp_start(expr, re_flags);
1010 regcode = JUST_CALC_SIZE;
1011 regc(REGMAGIC);
1012 if (reg(REG_NOPAREN, &flags) == NULL)
1013 return NULL;
1014
1015 /* Small enough for pointer-storage convention? */
1016 #ifdef SMALL_MALLOC /* 16 bit storage allocation */
1017 if (regsize >= 65536L - 256L)
1018 EMSG_RET_NULL(_("E339: Pattern too long"));
1019 #endif
1020
1021 /* Allocate space. */
1022 r = (regprog_T *)lalloc(sizeof(regprog_T) + regsize, TRUE);
1023 if (r == NULL)
1024 return NULL;
1025
1026 /*
1027 * Second pass: emit code.
1028 */
1029 regcomp_start(expr, re_flags);
1030 regcode = r->program;
1031 regc(REGMAGIC);
1032 if (reg(REG_NOPAREN, &flags) == NULL || reg_toolong)
1033 {
1034 vim_free(r);
1035 if (reg_toolong)
1036 EMSG_RET_NULL(_("E339: Pattern too long"));
1037 return NULL;
1038 }
1039
1040 /* Dig out information for optimizations. */
1041 r->regstart = NUL; /* Worst-case defaults. */
1042 r->reganch = 0;
1043 r->regmust = NULL;
1044 r->regmlen = 0;
1045 r->regflags = regflags;
1046 if (flags & HASNL)
1047 r->regflags |= RF_HASNL;
1048 if (flags & HASLOOKBH)
1049 r->regflags |= RF_LOOKBH;
1050 #ifdef FEAT_SYN_HL
1051 /* Remember whether this pattern has any \z specials in it. */
1052 r->reghasz = re_has_z;
1053 #endif
1054 scan = r->program + 1; /* First BRANCH. */
1055 if (OP(regnext(scan)) == END) /* Only one top-level choice. */
1056 {
1057 scan = OPERAND(scan);
1058
1059 /* Starting-point info. */
1060 if (OP(scan) == BOL || OP(scan) == RE_BOF)
1061 {
1062 r->reganch++;
1063 scan = regnext(scan);
1064 }
1065
1066 if (OP(scan) == EXACTLY)
1067 {
1068 #ifdef FEAT_MBYTE
1069 if (has_mbyte)
1070 r->regstart = (*mb_ptr2char)(OPERAND(scan));
1071 else
1072 #endif
1073 r->regstart = *OPERAND(scan);
1074 }
1075 else if ((OP(scan) == BOW
1076 || OP(scan) == EOW
1077 || OP(scan) == NOTHING
1078 || OP(scan) == MOPEN + 0 || OP(scan) == NOPEN
1079 || OP(scan) == MCLOSE + 0 || OP(scan) == NCLOSE)
1080 && OP(regnext(scan)) == EXACTLY)
1081 {
1082 #ifdef FEAT_MBYTE
1083 if (has_mbyte)
1084 r->regstart = (*mb_ptr2char)(OPERAND(regnext(scan)));
1085 else
1086 #endif
1087 r->regstart = *OPERAND(regnext(scan));
1088 }
1089
1090 /*
1091 * If there's something expensive in the r.e., find the longest
1092 * literal string that must appear and make it the regmust. Resolve
1093 * ties in favor of later strings, since the regstart check works
1094 * with the beginning of the r.e. and avoiding duplication
1095 * strengthens checking. Not a strong reason, but sufficient in the
1096 * absence of others.
1097 */
1098 /*
1099 * When the r.e. starts with BOW, it is faster to look for a regmust
1100 * first. Used a lot for "#" and "*" commands. (Added by mool).
1101 */
1102 if ((flags & SPSTART || OP(scan) == BOW || OP(scan) == EOW)
1103 && !(flags & HASNL))
1104 {
1105 longest = NULL;
1106 len = 0;
1107 for (; scan != NULL; scan = regnext(scan))
1108 if (OP(scan) == EXACTLY && STRLEN(OPERAND(scan)) >= (size_t)len)
1109 {
1110 longest = OPERAND(scan);
1111 len = (int)STRLEN(OPERAND(scan));
1112 }
1113 r->regmust = longest;
1114 r->regmlen = len;
1115 }
1116 }
1117 #ifdef DEBUG
1118 regdump(expr, r);
1119 #endif
1120 return r;
1121 }
1122
1123 /*
1124 * Setup to parse the regexp. Used once to get the length and once to do it.
1125 */
1126 static void
1127 regcomp_start(expr, re_flags)
1128 char_u *expr;
1129 int re_flags; /* see vim_regcomp() */
1130 {
1131 initchr(expr);
1132 if (re_flags & RE_MAGIC)
1133 reg_magic = MAGIC_ON;
1134 else
1135 reg_magic = MAGIC_OFF;
1136 reg_string = (re_flags & RE_STRING);
1137 reg_strict = (re_flags & RE_STRICT);
1138
1139 num_complex_braces = 0;
1140 regnpar = 1;
1141 vim_memset(had_endbrace, 0, sizeof(had_endbrace));
1142 #ifdef FEAT_SYN_HL
1143 regnzpar = 1;
1144 re_has_z = 0;
1145 #endif
1146 regsize = 0L;
1147 reg_toolong = FALSE;
1148 regflags = 0;
1149 #if defined(FEAT_SYN_HL) || defined(PROTO)
1150 had_eol = FALSE;
1151 #endif
1152 }
1153
1154 #if defined(FEAT_SYN_HL) || defined(PROTO)
1155 /*
1156 * Check if during the previous call to vim_regcomp the EOL item "$" has been
1157 * found. This is messy, but it works fine.
1158 */
1159 int
1160 vim_regcomp_had_eol()
1161 {
1162 return had_eol;
1163 }
1164 #endif
1165
1166 /*
1167 * reg - regular expression, i.e. main body or parenthesized thing
1168 *
1169 * Caller must absorb opening parenthesis.
1170 *
1171 * Combining parenthesis handling with the base level of regular expression
1172 * is a trifle forced, but the need to tie the tails of the branches to what
1173 * follows makes it hard to avoid.
1174 */
1175 static char_u *
1176 reg(paren, flagp)
1177 int paren; /* REG_NOPAREN, REG_PAREN, REG_NPAREN or REG_ZPAREN */
1178 int *flagp;
1179 {
1180 char_u *ret;
1181 char_u *br;
1182 char_u *ender;
1183 int parno = 0;
1184 int flags;
1185
1186 *flagp = HASWIDTH; /* Tentatively. */
1187
1188 #ifdef FEAT_SYN_HL
1189 if (paren == REG_ZPAREN)
1190 {
1191 /* Make a ZOPEN node. */
1192 if (regnzpar >= NSUBEXP)
1193 EMSG_RET_NULL(_("E50: Too many \\z("));
1194 parno = regnzpar;
1195 regnzpar++;
1196 ret = regnode(ZOPEN + parno);
1197 }
1198 else
1199 #endif
1200 if (paren == REG_PAREN)
1201 {
1202 /* Make a MOPEN node. */
1203 if (regnpar >= NSUBEXP)
1204 EMSG_M_RET_NULL(_("E51: Too many %s("), reg_magic == MAGIC_ALL);
1205 parno = regnpar;
1206 ++regnpar;
1207 ret = regnode(MOPEN + parno);
1208 }
1209 else if (paren == REG_NPAREN)
1210 {
1211 /* Make a NOPEN node. */
1212 ret = regnode(NOPEN);
1213 }
1214 else
1215 ret = NULL;
1216
1217 /* Pick up the branches, linking them together. */
1218 br = regbranch(&flags);
1219 if (br == NULL)
1220 return NULL;
1221 if (ret != NULL)
1222 regtail(ret, br); /* [MZ]OPEN -> first. */
1223 else
1224 ret = br;
1225 /* If one of the branches can be zero-width, the whole thing can.
1226 * If one of the branches has * at start or matches a line-break, the
1227 * whole thing can. */
1228 if (!(flags & HASWIDTH))
1229 *flagp &= ~HASWIDTH;
1230 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
1231 while (peekchr() == Magic('|'))
1232 {
1233 skipchr();
1234 br = regbranch(&flags);
1235 if (br == NULL || reg_toolong)
1236 return NULL;
1237 regtail(ret, br); /* BRANCH -> BRANCH. */
1238 if (!(flags & HASWIDTH))
1239 *flagp &= ~HASWIDTH;
1240 *flagp |= flags & (SPSTART | HASNL | HASLOOKBH);
1241 }
1242
1243 /* Make a closing node, and hook it on the end. */
1244 ender = regnode(
1245 #ifdef FEAT_SYN_HL
1246 paren == REG_ZPAREN ? ZCLOSE + parno :
1247 #endif
1248 paren == REG_PAREN ? MCLOSE + parno :
1249 paren == REG_NPAREN ? NCLOSE : END);
1250 regtail(ret, ender);
1251
1252 /* Hook the tails of the branches to the closing node. */
1253 for (br = ret; br != NULL; br = regnext(br))
1254 regoptail(br, ender);
1255
1256 /* Check for proper termination. */
1257 if (paren != REG_NOPAREN && getchr() != Magic(')'))
1258 {
1259 #ifdef FEAT_SYN_HL
1260 if (paren == REG_ZPAREN)
1261 EMSG_RET_NULL(_("E52: Unmatched \\z("));
1262 else
1263 #endif
1264 if (paren == REG_NPAREN)
1265 EMSG_M_RET_NULL(_("E53: Unmatched %s%%("), reg_magic == MAGIC_ALL);
1266 else
1267 EMSG_M_RET_NULL(_("E54: Unmatched %s("), reg_magic == MAGIC_ALL);
1268 }
1269 else if (paren == REG_NOPAREN && peekchr() != NUL)
1270 {
1271 if (curchr == Magic(')'))
1272 EMSG_M_RET_NULL(_("E55: Unmatched %s)"), reg_magic == MAGIC_ALL);
1273 else
1274 EMSG_RET_NULL(_(e_trailing)); /* "Can't happen". */
1275 /* NOTREACHED */
1276 }
1277 /*
1278 * Here we set the flag allowing back references to this set of
1279 * parentheses.
1280 */
1281 if (paren == REG_PAREN)
1282 had_endbrace[parno] = TRUE; /* have seen the close paren */
1283 return ret;
1284 }
1285
1286 /*
1287 * Handle one alternative of an | operator.
1288 * Implements the & operator.
1289 */
1290 static char_u *
1291 regbranch(flagp)
1292 int *flagp;
1293 {
1294 char_u *ret;
1295 char_u *chain = NULL;
1296 char_u *latest;
1297 int flags;
1298
1299 *flagp = WORST | HASNL; /* Tentatively. */
1300
1301 ret = regnode(BRANCH);
1302 for (;;)
1303 {
1304 latest = regconcat(&flags);
1305 if (latest == NULL)
1306 return NULL;
1307 /* If one of the branches has width, the whole thing has. If one of
1308 * the branches anchors at start-of-line, the whole thing does.
1309 * If one of the branches uses look-behind, the whole thing does. */
1310 *flagp |= flags & (HASWIDTH | SPSTART | HASLOOKBH);
1311 /* If one of the branches doesn't match a line-break, the whole thing
1312 * doesn't. */
1313 *flagp &= ~HASNL | (flags & HASNL);
1314 if (chain != NULL)
1315 regtail(chain, latest);
1316 if (peekchr() != Magic('&'))
1317 break;
1318 skipchr();
1319 regtail(latest, regnode(END)); /* operand ends */
1320 if (reg_toolong)
1321 break;
1322 reginsert(MATCH, latest);
1323 chain = latest;
1324 }
1325
1326 return ret;
1327 }
1328
1329 /*
1330 * Handle one alternative of an | or & operator.
1331 * Implements the concatenation operator.
1332 */
1333 static char_u *
1334 regconcat(flagp)
1335 int *flagp;
1336 {
1337 char_u *first = NULL;
1338 char_u *chain = NULL;
1339 char_u *latest;
1340 int flags;
1341 int cont = TRUE;
1342
1343 *flagp = WORST; /* Tentatively. */
1344
1345 while (cont)
1346 {
1347 switch (peekchr())
1348 {
1349 case NUL:
1350 case Magic('|'):
1351 case Magic('&'):
1352 case Magic(')'):
1353 cont = FALSE;
1354 break;
1355 case Magic('Z'):
1356 #ifdef FEAT_MBYTE
1357 regflags |= RF_ICOMBINE;
1358 #endif
1359 skipchr_keepstart();
1360 break;
1361 case Magic('c'):
1362 regflags |= RF_ICASE;
1363 skipchr_keepstart();
1364 break;
1365 case Magic('C'):
1366 regflags |= RF_NOICASE;
1367 skipchr_keepstart();
1368 break;
1369 case Magic('v'):
1370 reg_magic = MAGIC_ALL;
1371 skipchr_keepstart();
1372 curchr = -1;
1373 break;
1374 case Magic('m'):
1375 reg_magic = MAGIC_ON;
1376 skipchr_keepstart();
1377 curchr = -1;
1378 break;
1379 case Magic('M'):
1380 reg_magic = MAGIC_OFF;
1381 skipchr_keepstart();
1382 curchr = -1;
1383 break;
1384 case Magic('V'):
1385 reg_magic = MAGIC_NONE;
1386 skipchr_keepstart();
1387 curchr = -1;
1388 break;
1389 default:
1390 latest = regpiece(&flags);
1391 if (latest == NULL || reg_toolong)
1392 return NULL;
1393 *flagp |= flags & (HASWIDTH | HASNL | HASLOOKBH);
1394 if (chain == NULL) /* First piece. */
1395 *flagp |= flags & SPSTART;
1396 else
1397 regtail(chain, latest);
1398 chain = latest;
1399 if (first == NULL)
1400 first = latest;
1401 break;
1402 }
1403 }
1404 if (first == NULL) /* Loop ran zero times. */
1405 first = regnode(NOTHING);
1406 return first;
1407 }
1408
1409 /*
1410 * regpiece - something followed by possible [*+=]
1411 *
1412 * Note that the branching code sequences used for = and the general cases
1413 * of * and + are somewhat optimized: they use the same NOTHING node as
1414 * both the endmarker for their branch list and the body of the last branch.
1415 * It might seem that this node could be dispensed with entirely, but the
1416 * endmarker role is not redundant.
1417 */
1418 static char_u *
1419 regpiece(flagp)
1420 int *flagp;
1421 {
1422 char_u *ret;
1423 int op;
1424 char_u *next;
1425 int flags;
1426 long minval;
1427 long maxval;
1428
1429 ret = regatom(&flags);
1430 if (ret == NULL)
1431 return NULL;
1432
1433 op = peekchr();
1434 if (re_multi_type(op) == NOT_MULTI)
1435 {
1436 *flagp = flags;
1437 return ret;
1438 }
1439 /* default flags */
1440 *flagp = (WORST | SPSTART | (flags & (HASNL | HASLOOKBH)));
1441
1442 skipchr();
1443 switch (op)
1444 {
1445 case Magic('*'):
1446 if (flags & SIMPLE)
1447 reginsert(STAR, ret);
1448 else
1449 {
1450 /* Emit x* as (x&|), where & means "self". */
1451 reginsert(BRANCH, ret); /* Either x */
1452 regoptail(ret, regnode(BACK)); /* and loop */
1453 regoptail(ret, ret); /* back */
1454 regtail(ret, regnode(BRANCH)); /* or */
1455 regtail(ret, regnode(NOTHING)); /* null. */
1456 }
1457 break;
1458
1459 case Magic('+'):
1460 if (flags & SIMPLE)
1461 reginsert(PLUS, ret);
1462 else
1463 {
1464 /* Emit x+ as x(&|), where & means "self". */
1465 next = regnode(BRANCH); /* Either */
1466 regtail(ret, next);
1467 regtail(regnode(BACK), ret); /* loop back */
1468 regtail(next, regnode(BRANCH)); /* or */
1469 regtail(ret, regnode(NOTHING)); /* null. */
1470 }
1471 *flagp = (WORST | HASWIDTH | (flags & (HASNL | HASLOOKBH)));
1472 break;
1473
1474 case Magic('@'):
1475 {
1476 int lop = END;
1477
1478 switch (no_Magic(getchr()))
1479 {
1480 case '=': lop = MATCH; break; /* \@= */
1481 case '!': lop = NOMATCH; break; /* \@! */
1482 case '>': lop = SUBPAT; break; /* \@> */
1483 case '<': switch (no_Magic(getchr()))
1484 {
1485 case '=': lop = BEHIND; break; /* \@<= */
1486 case '!': lop = NOBEHIND; break; /* \@<! */
1487 }
1488 }
1489 if (lop == END)
1490 EMSG_M_RET_NULL(_("E59: invalid character after %s@"),
1491 reg_magic == MAGIC_ALL);
1492 /* Look behind must match with behind_pos. */
1493 if (lop == BEHIND || lop == NOBEHIND)
1494 {
1495 regtail(ret, regnode(BHPOS));
1496 *flagp |= HASLOOKBH;
1497 }
1498 regtail(ret, regnode(END)); /* operand ends */
1499 reginsert(lop, ret);
1500 break;
1501 }
1502
1503 case Magic('?'):
1504 case Magic('='):
1505 /* Emit x= as (x|) */
1506 reginsert(BRANCH, ret); /* Either x */
1507 regtail(ret, regnode(BRANCH)); /* or */
1508 next = regnode(NOTHING); /* null. */
1509 regtail(ret, next);
1510 regoptail(ret, next);
1511 break;
1512
1513 case Magic('{'):
1514 if (!read_limits(&minval, &maxval))
1515 return NULL;
1516 if (flags & SIMPLE)
1517 {
1518 reginsert(BRACE_SIMPLE, ret);
1519 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
1520 }
1521 else
1522 {
1523 if (num_complex_braces >= 10)
1524 EMSG_M_RET_NULL(_("E60: Too many complex %s{...}s"),
1525 reg_magic == MAGIC_ALL);
1526 reginsert(BRACE_COMPLEX + num_complex_braces, ret);
1527 regoptail(ret, regnode(BACK));
1528 regoptail(ret, ret);
1529 reginsert_limits(BRACE_LIMITS, minval, maxval, ret);
1530 ++num_complex_braces;
1531 }
1532 if (minval > 0 && maxval > 0)
1533 *flagp = (HASWIDTH | (flags & (HASNL | HASLOOKBH)));
1534 break;
1535 }
1536 if (re_multi_type(peekchr()) != NOT_MULTI)
1537 {
1538 /* Can't have a multi follow a multi. */
1539 if (peekchr() == Magic('*'))
1540 sprintf((char *)IObuff, _("E61: Nested %s*"),
1541 reg_magic >= MAGIC_ON ? "" : "\\");
1542 else
1543 sprintf((char *)IObuff, _("E62: Nested %s%c"),
1544 reg_magic == MAGIC_ALL ? "" : "\\", no_Magic(peekchr()));
1545 EMSG_RET_NULL(IObuff);
1546 }
1547
1548 return ret;
1549 }
1550
1551 /*
1552 * regatom - the lowest level
1553 *
1554 * Optimization: gobbles an entire sequence of ordinary characters so that
1555 * it can turn them into a single node, which is smaller to store and
1556 * faster to run. Don't do this when one_exactly is set.
1557 */
1558 static char_u *
1559 regatom(flagp)
1560 int *flagp;
1561 {
1562 char_u *ret;
1563 int flags;
1564 int cpo_lit; /* 'cpoptions' contains 'l' flag */
1565 int cpo_bsl; /* 'cpoptions' contains '\' flag */
1566 int c;
1567 static char_u *classchars = (char_u *)".iIkKfFpPsSdDxXoOwWhHaAlLuU";
1568 static int classcodes[] = {ANY, IDENT, SIDENT, KWORD, SKWORD,
1569 FNAME, SFNAME, PRINT, SPRINT,
1570 WHITE, NWHITE, DIGIT, NDIGIT,
1571 HEX, NHEX, OCTAL, NOCTAL,
1572 WORD, NWORD, HEAD, NHEAD,
1573 ALPHA, NALPHA, LOWER, NLOWER,
1574 UPPER, NUPPER
1575 };
1576 char_u *p;
1577 int extra = 0;
1578
1579 *flagp = WORST; /* Tentatively. */
1580 cpo_lit = vim_strchr(p_cpo, CPO_LITERAL) != NULL;
1581 cpo_bsl = vim_strchr(p_cpo, CPO_BACKSL) != NULL;
1582
1583 c = getchr();
1584 switch (c)
1585 {
1586 case Magic('^'):
1587 ret = regnode(BOL);
1588 break;
1589
1590 case Magic('$'):
1591 ret = regnode(EOL);
1592 #if defined(FEAT_SYN_HL) || defined(PROTO)
1593 had_eol = TRUE;
1594 #endif
1595 break;
1596
1597 case Magic('<'):
1598 ret = regnode(BOW);
1599 break;
1600
1601 case Magic('>'):
1602 ret = regnode(EOW);
1603 break;
1604
1605 case Magic('_'):
1606 c = no_Magic(getchr());
1607 if (c == '^') /* "\_^" is start-of-line */
1608 {
1609 ret = regnode(BOL);
1610 break;
1611 }
1612 if (c == '$') /* "\_$" is end-of-line */
1613 {
1614 ret = regnode(EOL);
1615 #if defined(FEAT_SYN_HL) || defined(PROTO)
1616 had_eol = TRUE;
1617 #endif
1618 break;
1619 }
1620
1621 extra = ADD_NL;
1622 *flagp |= HASNL;
1623
1624 /* "\_[" is character range plus newline */
1625 if (c == '[')
1626 goto collection;
1627
1628 /* "\_x" is character class plus newline */
1629 /*FALLTHROUGH*/
1630
1631 /*
1632 * Character classes.
1633 */
1634 case Magic('.'):
1635 case Magic('i'):
1636 case Magic('I'):
1637 case Magic('k'):
1638 case Magic('K'):
1639 case Magic('f'):
1640 case Magic('F'):
1641 case Magic('p'):
1642 case Magic('P'):
1643 case Magic('s'):
1644 case Magic('S'):
1645 case Magic('d'):
1646 case Magic('D'):
1647 case Magic('x'):
1648 case Magic('X'):
1649 case Magic('o'):
1650 case Magic('O'):
1651 case Magic('w'):
1652 case Magic('W'):
1653 case Magic('h'):
1654 case Magic('H'):
1655 case Magic('a'):
1656 case Magic('A'):
1657 case Magic('l'):
1658 case Magic('L'):
1659 case Magic('u'):
1660 case Magic('U'):
1661 p = vim_strchr(classchars, no_Magic(c));
1662 if (p == NULL)
1663 EMSG_RET_NULL(_("E63: invalid use of \\_"));
1664 #ifdef FEAT_MBYTE
1665 /* When '.' is followed by a composing char ignore the dot, so that
1666 * the composing char is matched here. */
1667 if (enc_utf8 && c == Magic('.') && utf_iscomposing(peekchr()))
1668 {
1669 c = getchr();
1670 goto do_multibyte;
1671 }
1672 #endif
1673 ret = regnode(classcodes[p - classchars] + extra);
1674 *flagp |= HASWIDTH | SIMPLE;
1675 break;
1676
1677 case Magic('n'):
1678 if (reg_string)
1679 {
1680 /* In a string "\n" matches a newline character. */
1681 ret = regnode(EXACTLY);
1682 regc(NL);
1683 regc(NUL);
1684 *flagp |= HASWIDTH | SIMPLE;
1685 }
1686 else
1687 {
1688 /* In buffer text "\n" matches the end of a line. */
1689 ret = regnode(NEWL);
1690 *flagp |= HASWIDTH | HASNL;
1691 }
1692 break;
1693
1694 case Magic('('):
1695 if (one_exactly)
1696 EMSG_ONE_RET_NULL;
1697 ret = reg(REG_PAREN, &flags);
1698 if (ret == NULL)
1699 return NULL;
1700 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
1701 break;
1702
1703 case NUL:
1704 case Magic('|'):
1705 case Magic('&'):
1706 case Magic(')'):
1707 if (one_exactly)
1708 EMSG_ONE_RET_NULL;
1709 EMSG_RET_NULL(_(e_internal)); /* Supposed to be caught earlier. */
1710 /* NOTREACHED */
1711
1712 case Magic('='):
1713 case Magic('?'):
1714 case Magic('+'):
1715 case Magic('@'):
1716 case Magic('{'):
1717 case Magic('*'):
1718 c = no_Magic(c);
1719 sprintf((char *)IObuff, _("E64: %s%c follows nothing"),
1720 (c == '*' ? reg_magic >= MAGIC_ON : reg_magic == MAGIC_ALL)
1721 ? "" : "\\", c);
1722 EMSG_RET_NULL(IObuff);
1723 /* NOTREACHED */
1724
1725 case Magic('~'): /* previous substitute pattern */
1726 if (reg_prev_sub != NULL)
1727 {
1728 char_u *lp;
1729
1730 ret = regnode(EXACTLY);
1731 lp = reg_prev_sub;
1732 while (*lp != NUL)
1733 regc(*lp++);
1734 regc(NUL);
1735 if (*reg_prev_sub != NUL)
1736 {
1737 *flagp |= HASWIDTH;
1738 if ((lp - reg_prev_sub) == 1)
1739 *flagp |= SIMPLE;
1740 }
1741 }
1742 else
1743 EMSG_RET_NULL(_(e_nopresub));
1744 break;
1745
1746 case Magic('1'):
1747 case Magic('2'):
1748 case Magic('3'):
1749 case Magic('4'):
1750 case Magic('5'):
1751 case Magic('6'):
1752 case Magic('7'):
1753 case Magic('8'):
1754 case Magic('9'):
1755 {
1756 int refnum;
1757
1758 refnum = c - Magic('0');
1759 /*
1760 * Check if the back reference is legal. We must have seen the
1761 * close brace.
1762 * TODO: Should also check that we don't refer to something
1763 * that is repeated (+*=): what instance of the repetition
1764 * should we match?
1765 */
1766 if (!had_endbrace[refnum])
1767 {
1768 /* Trick: check if "@<=" or "@<!" follows, in which case
1769 * the \1 can appear before the referenced match. */
1770 for (p = regparse; *p != NUL; ++p)
1771 if (p[0] == '@' && p[1] == '<'
1772 && (p[2] == '!' || p[2] == '='))
1773 break;
1774 if (*p == NUL)
1775 EMSG_RET_NULL(_("E65: Illegal back reference"));
1776 }
1777 ret = regnode(BACKREF + refnum);
1778 }
1779 break;
1780
1781 case Magic('z'):
1782 {
1783 c = no_Magic(getchr());
1784 switch (c)
1785 {
1786 #ifdef FEAT_SYN_HL
1787 case '(': if (reg_do_extmatch != REX_SET)
1788 EMSG_RET_NULL(_("E66: \\z( not allowed here"));
1789 if (one_exactly)
1790 EMSG_ONE_RET_NULL;
1791 ret = reg(REG_ZPAREN, &flags);
1792 if (ret == NULL)
1793 return NULL;
1794 *flagp |= flags & (HASWIDTH|SPSTART|HASNL|HASLOOKBH);
1795 re_has_z = REX_SET;
1796 break;
1797
1798 case '1':
1799 case '2':
1800 case '3':
1801 case '4':
1802 case '5':
1803 case '6':
1804 case '7':
1805 case '8':
1806 case '9': if (reg_do_extmatch != REX_USE)
1807 EMSG_RET_NULL(_("E67: \\z1 et al. not allowed here"));
1808 ret = regnode(ZREF + c - '0');
1809 re_has_z = REX_USE;
1810 break;
1811 #endif
1812
1813 case 's': ret = regnode(MOPEN + 0);
1814 break;
1815
1816 case 'e': ret = regnode(MCLOSE + 0);
1817 break;
1818
1819 default: EMSG_RET_NULL(_("E68: Invalid character after \\z"));
1820 }
1821 }
1822 break;
1823
1824 case Magic('%'):
1825 {
1826 c = no_Magic(getchr());
1827 switch (c)
1828 {
1829 /* () without a back reference */
1830 case '(':
1831 if (one_exactly)
1832 EMSG_ONE_RET_NULL;
1833 ret = reg(REG_NPAREN, &flags);
1834 if (ret == NULL)
1835 return NULL;
1836 *flagp |= flags & (HASWIDTH | SPSTART | HASNL | HASLOOKBH);
1837 break;
1838
1839 /* Catch \%^ and \%$ regardless of where they appear in the
1840 * pattern -- regardless of whether or not it makes sense. */
1841 case '^':
1842 ret = regnode(RE_BOF);
1843 break;
1844
1845 case '$':
1846 ret = regnode(RE_EOF);
1847 break;
1848
1849 case '#':
1850 ret = regnode(CURSOR);
1851 break;
1852
1853 case 'V':
1854 ret = regnode(RE_VISUAL);
1855 break;
1856
1857 /* \%[abc]: Emit as a list of branches, all ending at the last
1858 * branch which matches nothing. */
1859 case '[':
1860 if (one_exactly) /* doesn't nest */
1861 EMSG_ONE_RET_NULL;
1862 {
1863 char_u *lastbranch;
1864 char_u *lastnode = NULL;
1865 char_u *br;
1866
1867 ret = NULL;
1868 while ((c = getchr()) != ']')
1869 {
1870 if (c == NUL)
1871 EMSG_M_RET_NULL(_("E69: Missing ] after %s%%["),
1872 reg_magic == MAGIC_ALL);
1873 br = regnode(BRANCH);
1874 if (ret == NULL)
1875 ret = br;
1876 else
1877 regtail(lastnode, br);
1878
1879 ungetchr();
1880 one_exactly = TRUE;
1881 lastnode = regatom(flagp);
1882 one_exactly = FALSE;
1883 if (lastnode == NULL)
1884 return NULL;
1885 }
1886 if (ret == NULL)
1887 EMSG_M_RET_NULL(_("E70: Empty %s%%[]"),
1888 reg_magic == MAGIC_ALL);
1889 lastbranch = regnode(BRANCH);
1890 br = regnode(NOTHING);
1891 if (ret != JUST_CALC_SIZE)
1892 {
1893 regtail(lastnode, br);
1894 regtail(lastbranch, br);
1895 /* connect all branches to the NOTHING
1896 * branch at the end */
1897 for (br = ret; br != lastnode; )
1898 {
1899 if (OP(br) == BRANCH)
1900 {
1901 regtail(br, lastbranch);
1902 br = OPERAND(br);
1903 }
1904 else
1905 br = regnext(br);
1906 }
1907 }
1908 *flagp &= ~(HASWIDTH | SIMPLE);
1909 break;
1910 }
1911
1912 case 'd': /* %d123 decimal */
1913 case 'o': /* %o123 octal */
1914 case 'x': /* %xab hex 2 */
1915 case 'u': /* %uabcd hex 4 */
1916 case 'U': /* %U1234abcd hex 8 */
1917 {
1918 int i;
1919
1920 switch (c)
1921 {
1922 case 'd': i = getdecchrs(); break;
1923 case 'o': i = getoctchrs(); break;
1924 case 'x': i = gethexchrs(2); break;
1925 case 'u': i = gethexchrs(4); break;
1926 case 'U': i = gethexchrs(8); break;
1927 default: i = -1; break;
1928 }
1929
1930 if (i < 0)
1931 EMSG_M_RET_NULL(
1932 _("E678: Invalid character after %s%%[dxouU]"),
1933 reg_magic == MAGIC_ALL);
1934 #ifdef FEAT_MBYTE
1935 if (use_multibytecode(i))
1936 ret = regnode(MULTIBYTECODE);
1937 else
1938 #endif
1939 ret = regnode(EXACTLY);
1940 if (i == 0)
1941 regc(0x0a);
1942 else
1943 #ifdef FEAT_MBYTE
1944 regmbc(i);
1945 #else
1946 regc(i);
1947 #endif
1948 regc(NUL);
1949 *flagp |= HASWIDTH;
1950 break;
1951 }
1952
1953 default:
1954 if (VIM_ISDIGIT(c) || c == '<' || c == '>'
1955 || c == '\'')
1956 {
1957 long_u n = 0;
1958 int cmp;
1959
1960 cmp = c;
1961 if (cmp == '<' || cmp == '>')
1962 c = getchr();
1963 while (VIM_ISDIGIT(c))
1964 {
1965 n = n * 10 + (c - '0');
1966 c = getchr();
1967 }
1968 if (c == '\'' && n == 0)
1969 {
1970 /* "\%'m", "\%<'m" and "\%>'m": Mark */
1971 c = getchr();
1972 ret = regnode(RE_MARK);
1973 if (ret == JUST_CALC_SIZE)
1974 regsize += 2;
1975 else
1976 {
1977 *regcode++ = c;
1978 *regcode++ = cmp;
1979 }
1980 break;
1981 }
1982 else if (c == 'l' || c == 'c' || c == 'v')
1983 {
1984 if (c == 'l')
1985 ret = regnode(RE_LNUM);
1986 else if (c == 'c')
1987 ret = regnode(RE_COL);
1988 else
1989 ret = regnode(RE_VCOL);
1990 if (ret == JUST_CALC_SIZE)
1991 regsize += 5;
1992 else
1993 {
1994 /* put the number and the optional
1995 * comparator after the opcode */
1996 regcode = re_put_long(regcode, n);
1997 *regcode++ = cmp;
1998 }
1999 break;
2000 }
2001 }
2002
2003 EMSG_M_RET_NULL(_("E71: Invalid character after %s%%"),
2004 reg_magic == MAGIC_ALL);
2005 }
2006 }
2007 break;
2008
2009 case Magic('['):
2010 collection:
2011 {
2012 char_u *lp;
2013
2014 /*
2015 * If there is no matching ']', we assume the '[' is a normal
2016 * character. This makes 'incsearch' and ":help [" work.
2017 */
2018 lp = skip_anyof(regparse);
2019 if (*lp == ']') /* there is a matching ']' */
2020 {
2021 int startc = -1; /* > 0 when next '-' is a range */
2022 int endc;
2023
2024 /*
2025 * In a character class, different parsing rules apply.
2026 * Not even \ is special anymore, nothing is.
2027 */
2028 if (*regparse == '^') /* Complement of range. */
2029 {
2030 ret = regnode(ANYBUT + extra);
2031 regparse++;
2032 }
2033 else
2034 ret = regnode(ANYOF + extra);
2035
2036 /* At the start ']' and '-' mean the literal character. */
2037 if (*regparse == ']' || *regparse == '-')
2038 {
2039 startc = *regparse;
2040 regc(*regparse++);
2041 }
2042
2043 while (*regparse != NUL && *regparse != ']')
2044 {
2045 if (*regparse == '-')
2046 {
2047 ++regparse;
2048 /* The '-' is not used for a range at the end and
2049 * after or before a '\n'. */
2050 if (*regparse == ']' || *regparse == NUL
2051 || startc == -1
2052 || (regparse[0] == '\\' && regparse[1] == 'n'))
2053 {
2054 regc('-');
2055 startc = '-'; /* [--x] is a range */
2056 }
2057 else
2058 {
2059 /* Also accept "a-[.z.]" */
2060 endc = 0;
2061 if (*regparse == '[')
2062 endc = get_coll_element(&regparse);
2063 if (endc == 0)
2064 {
2065 #ifdef FEAT_MBYTE
2066 if (has_mbyte)
2067 endc = mb_ptr2char_adv(&regparse);
2068 else
2069 #endif
2070 endc = *regparse++;
2071 }
2072
2073 /* Handle \o40, \x20 and \u20AC style sequences */
2074 if (endc == '\\' && !cpo_lit && !cpo_bsl)
2075 endc = coll_get_char();
2076
2077 if (startc > endc)
2078 EMSG_RET_NULL(_(e_invrange));
2079 #ifdef FEAT_MBYTE
2080 if (has_mbyte && ((*mb_char2len)(startc) > 1
2081 || (*mb_char2len)(endc) > 1))
2082 {
2083 /* Limit to a range of 256 chars */
2084 if (endc > startc + 256)
2085 EMSG_RET_NULL(_(e_invrange));
2086 while (++startc <= endc)
2087 regmbc(startc);
2088 }
2089 else
2090 #endif
2091 {
2092 #ifdef EBCDIC
2093 int alpha_only = FALSE;
2094
2095 /* for alphabetical range skip the gaps
2096 * 'i'-'j', 'r'-'s', 'I'-'J' and 'R'-'S'. */
2097 if (isalpha(startc) && isalpha(endc))
2098 alpha_only = TRUE;
2099 #endif
2100 while (++startc <= endc)
2101 #ifdef EBCDIC
2102 if (!alpha_only || isalpha(startc))
2103 #endif
2104 regc(startc);
2105 }
2106 startc = -1;
2107 }
2108 }
2109 /*
2110 * Only "\]", "\^", "\]" and "\\" are special in Vi. Vim
2111 * accepts "\t", "\e", etc., but only when the 'l' flag in
2112 * 'cpoptions' is not included.
2113 * Posix doesn't recognize backslash at all.
2114 */
2115 else if (*regparse == '\\'
2116 && !cpo_bsl
2117 && (vim_strchr(REGEXP_INRANGE, regparse[1]) != NULL
2118 || (!cpo_lit
2119 && vim_strchr(REGEXP_ABBR,
2120 regparse[1]) != NULL)))
2121 {
2122 regparse++;
2123 if (*regparse == 'n')
2124 {
2125 /* '\n' in range: also match NL */
2126 if (ret != JUST_CALC_SIZE)
2127 {
2128 if (*ret == ANYBUT)
2129 *ret = ANYBUT + ADD_NL;
2130 else if (*ret == ANYOF)
2131 *ret = ANYOF + ADD_NL;
2132 /* else: must have had a \n already */
2133 }
2134 *flagp |= HASNL;
2135 regparse++;
2136 startc = -1;
2137 }
2138 else if (*regparse == 'd'
2139 || *regparse == 'o'
2140 || *regparse == 'x'
2141 || *regparse == 'u'
2142 || *regparse == 'U')
2143 {
2144 startc = coll_get_char();
2145 if (startc == 0)
2146 regc(0x0a);
2147 else
2148 #ifdef FEAT_MBYTE
2149 regmbc(startc);
2150 #else
2151 regc(startc);
2152 #endif
2153 }
2154 else
2155 {
2156 startc = backslash_trans(*regparse++);
2157 regc(startc);
2158 }
2159 }
2160 else if (*regparse == '[')
2161 {
2162 int c_class;
2163 int cu;
2164
2165 c_class = get_char_class(&regparse);
2166 startc = -1;
2167 /* Characters assumed to be 8 bits! */
2168 switch (c_class)
2169 {
2170 case CLASS_NONE:
2171 c_class = get_equi_class(&regparse);
2172 if (c_class != 0)
2173 {
2174 /* produce equivalence class */
2175 reg_equi_class(c_class);
2176 }
2177 else if ((c_class =
2178 get_coll_element(&regparse)) != 0)
2179 {
2180 /* produce a collating element */
2181 regmbc(c_class);
2182 }
2183 else
2184 {
2185 /* literal '[', allow [[-x] as a range */
2186 startc = *regparse++;
2187 regc(startc);
2188 }
2189 break;
2190 case CLASS_ALNUM:
2191 for (cu = 1; cu <= 255; cu++)
2192 if (isalnum(cu))
2193 regc(cu);
2194 break;
2195 case CLASS_ALPHA:
2196 for (cu = 1; cu <= 255; cu++)
2197 if (isalpha(cu))
2198 regc(cu);
2199 break;
2200 case CLASS_BLANK:
2201 regc(' ');
2202 regc('\t');
2203 break;
2204 case CLASS_CNTRL:
2205 for (cu = 1; cu <= 255; cu++)
2206 if (iscntrl(cu))
2207 regc(cu);
2208 break;
2209 case CLASS_DIGIT:
2210 for (cu = 1; cu <= 255; cu++)
2211 if (VIM_ISDIGIT(cu))
2212 regc(cu);
2213 break;
2214 case CLASS_GRAPH:
2215 for (cu = 1; cu <= 255; cu++)
2216 if (isgraph(cu))
2217 regc(cu);
2218 break;
2219 case CLASS_LOWER:
2220 for (cu = 1; cu <= 255; cu++)
2221 if (MB_ISLOWER(cu))
2222 regc(cu);
2223 break;
2224 case CLASS_PRINT:
2225 for (cu = 1; cu <= 255; cu++)
2226 if (vim_isprintc(cu))
2227 regc(cu);
2228 break;
2229 case CLASS_PUNCT:
2230 for (cu = 1; cu <= 255; cu++)
2231 if (ispunct(cu))
2232 regc(cu);
2233 break;
2234 case CLASS_SPACE:
2235 for (cu = 9; cu <= 13; cu++)
2236 regc(cu);
2237 regc(' ');
2238 break;
2239 case CLASS_UPPER:
2240 for (cu = 1; cu <= 255; cu++)
2241 if (MB_ISUPPER(cu))
2242 regc(cu);
2243 break;
2244 case CLASS_XDIGIT:
2245 for (cu = 1; cu <= 255; cu++)
2246 if (vim_isxdigit(cu))
2247 regc(cu);
2248 break;
2249 case CLASS_TAB:
2250 regc('\t');
2251 break;
2252 case CLASS_RETURN:
2253 regc('\r');
2254 break;
2255 case CLASS_BACKSPACE:
2256 regc('\b');
2257 break;
2258 case CLASS_ESCAPE:
2259 regc('\033');
2260 break;
2261 }
2262 }
2263 else
2264 {
2265 #ifdef FEAT_MBYTE
2266 if (has_mbyte)
2267 {
2268 int len;
2269
2270 /* produce a multibyte character, including any
2271 * following composing characters */
2272 startc = mb_ptr2char(regparse);
2273 len = (*mb_ptr2len)(regparse);
2274 if (enc_utf8 && utf_char2len(startc) != len)
2275 startc = -1; /* composing chars */
2276 while (--len >= 0)
2277 regc(*regparse++);
2278 }
2279 else
2280 #endif
2281 {
2282 startc = *regparse++;
2283 regc(startc);
2284 }
2285 }
2286 }
2287 regc(NUL);
2288 prevchr_len = 1; /* last char was the ']' */
2289 if (*regparse != ']')
2290 EMSG_RET_NULL(_(e_toomsbra)); /* Cannot happen? */
2291 skipchr(); /* let's be friends with the lexer again */
2292 *flagp |= HASWIDTH | SIMPLE;
2293 break;
2294 }
2295 else if (reg_strict)
2296 EMSG_M_RET_NULL(_("E769: Missing ] after %s["),
2297 reg_magic > MAGIC_OFF);
2298 }
2299 /* FALLTHROUGH */
2300
2301 default:
2302 {
2303 int len;
2304
2305 #ifdef FEAT_MBYTE
2306 /* A multi-byte character is handled as a separate atom if it's
2307 * before a multi and when it's a composing char. */
2308 if (use_multibytecode(c))
2309 {
2310 do_multibyte:
2311 ret = regnode(MULTIBYTECODE);
2312 regmbc(c);
2313 *flagp |= HASWIDTH | SIMPLE;
2314 break;
2315 }
2316 #endif
2317
2318 ret = regnode(EXACTLY);
2319
2320 /*
2321 * Append characters as long as:
2322 * - there is no following multi, we then need the character in
2323 * front of it as a single character operand
2324 * - not running into a Magic character
2325 * - "one_exactly" is not set
2326 * But always emit at least one character. Might be a Multi,
2327 * e.g., a "[" without matching "]".
2328 */
2329 for (len = 0; c != NUL && (len == 0
2330 || (re_multi_type(peekchr()) == NOT_MULTI
2331 && !one_exactly
2332 && !is_Magic(c))); ++len)
2333 {
2334 c = no_Magic(c);
2335 #ifdef FEAT_MBYTE
2336 if (has_mbyte)
2337 {
2338 regmbc(c);
2339 if (enc_utf8)
2340 {
2341 int l;
2342
2343 /* Need to get composing character too. */
2344 for (;;)
2345 {
2346 l = utf_ptr2len(regparse);
2347 if (!UTF_COMPOSINGLIKE(regparse, regparse + l))
2348 break;
2349 regmbc(utf_ptr2char(regparse));
2350 skipchr();
2351 }
2352 }
2353 }
2354 else
2355 #endif
2356 regc(c);
2357 c = getchr();
2358 }
2359 ungetchr();
2360
2361 regc(NUL);
2362 *flagp |= HASWIDTH;
2363 if (len == 1)
2364 *flagp |= SIMPLE;
2365 }
2366 break;
2367 }
2368
2369 return ret;
2370 }
2371
2372 #ifdef FEAT_MBYTE
2373 /*
2374 * Return TRUE if MULTIBYTECODE should be used instead of EXACTLY for
2375 * character "c".
2376 */
2377 static int
2378 use_multibytecode(c)
2379 int c;
2380 {
2381 return has_mbyte && (*mb_char2len)(c) > 1
2382 && (re_multi_type(peekchr()) != NOT_MULTI
2383 || (enc_utf8 && utf_iscomposing(c)));
2384 }
2385 #endif
2386
2387 /*
2388 * emit a node
2389 * Return pointer to generated code.
2390 */
2391 static char_u *
2392 regnode(op)
2393 int op;
2394 {
2395 char_u *ret;
2396
2397 ret = regcode;
2398 if (ret == JUST_CALC_SIZE)
2399 regsize += 3;
2400 else
2401 {
2402 *regcode++ = op;
2403 *regcode++ = NUL; /* Null "next" pointer. */
2404 *regcode++ = NUL;
2405 }
2406 return ret;
2407 }
2408
2409 /*
2410 * Emit (if appropriate) a byte of code
2411 */
2412 static void
2413 regc(b)
2414 int b;
2415 {
2416 if (regcode == JUST_CALC_SIZE)
2417 regsize++;
2418 else
2419 *regcode++ = b;
2420 }
2421
2422 #ifdef FEAT_MBYTE
2423 /*
2424 * Emit (if appropriate) a multi-byte character of code
2425 */
2426 static void
2427 regmbc(c)
2428 int c;
2429 {
2430 if (regcode == JUST_CALC_SIZE)
2431 regsize += (*mb_char2len)(c);
2432 else
2433 regcode += (*mb_char2bytes)(c, regcode);
2434 }
2435 #endif
2436
2437 /*
2438 * reginsert - insert an operator in front of already-emitted operand
2439 *
2440 * Means relocating the operand.
2441 */
2442 static void
2443 reginsert(op, opnd)
2444 int op;
2445 char_u *opnd;
2446 {
2447 char_u *src;
2448 char_u *dst;
2449 char_u *place;
2450
2451 if (regcode == JUST_CALC_SIZE)
2452 {
2453 regsize += 3;
2454 return;
2455 }
2456 src = regcode;
2457 regcode += 3;
2458 dst = regcode;
2459 while (src > opnd)
2460 *--dst = *--src;
2461
2462 place = opnd; /* Op node, where operand used to be. */
2463 *place++ = op;
2464 *place++ = NUL;
2465 *place = NUL;
2466 }
2467
2468 /*
2469 * reginsert_limits - insert an operator in front of already-emitted operand.
2470 * The operator has the given limit values as operands. Also set next pointer.
2471 *
2472 * Means relocating the operand.
2473 */
2474 static void
2475 reginsert_limits(op, minval, maxval, opnd)
2476 int op;
2477 long minval;
2478 long maxval;
2479 char_u *opnd;
2480 {
2481 char_u *src;
2482 char_u *dst;
2483 char_u *place;
2484
2485 if (regcode == JUST_CALC_SIZE)
2486 {
2487 regsize += 11;
2488 return;
2489 }
2490 src = regcode;
2491 regcode += 11;
2492 dst = regcode;
2493 while (src > opnd)
2494 *--dst = *--src;
2495
2496 place = opnd; /* Op node, where operand used to be. */
2497 *place++ = op;
2498 *place++ = NUL;
2499 *place++ = NUL;
2500 place = re_put_long(place, (long_u)minval);
2501 place = re_put_long(place, (long_u)maxval);
2502 regtail(opnd, place);
2503 }
2504
2505 /*
2506 * Write a long as four bytes at "p" and return pointer to the next char.
2507 */
2508 static char_u *
2509 re_put_long(p, val)
2510 char_u *p;
2511 long_u val;
2512 {
2513 *p++ = (char_u) ((val >> 24) & 0377);
2514 *p++ = (char_u) ((val >> 16) & 0377);
2515 *p++ = (char_u) ((val >> 8) & 0377);
2516 *p++ = (char_u) (val & 0377);
2517 return p;
2518 }
2519
2520 /*
2521 * regtail - set the next-pointer at the end of a node chain
2522 */
2523 static void
2524 regtail(p, val)
2525 char_u *p;
2526 char_u *val;
2527 {
2528 char_u *scan;
2529 char_u *temp;
2530 int offset;
2531
2532 if (p == JUST_CALC_SIZE)
2533 return;
2534
2535 /* Find last node. */
2536 scan = p;
2537 for (;;)
2538 {
2539 temp = regnext(scan);
2540 if (temp == NULL)
2541 break;
2542 scan = temp;
2543 }
2544
2545 if (OP(scan) == BACK)
2546 offset = (int)(scan - val);
2547 else
2548 offset = (int)(val - scan);
2549 /* When the offset uses more than 16 bits it can no longer fit in the two
2550 * bytes avaliable. Use a global flag to avoid having to check return
2551 * values in too many places. */
2552 if (offset > 0xffff)
2553 reg_toolong = TRUE;
2554 else
2555 {
2556 *(scan + 1) = (char_u) (((unsigned)offset >> 8) & 0377);
2557 *(scan + 2) = (char_u) (offset & 0377);
2558 }
2559 }
2560
2561 /*
2562 * regoptail - regtail on item after a BRANCH; nop if none
2563 */
2564 static void
2565 regoptail(p, val)
2566 char_u *p;
2567 char_u *val;
2568 {
2569 /* When op is neither BRANCH nor BRACE_COMPLEX0-9, it is "operandless" */
2570 if (p == NULL || p == JUST_CALC_SIZE
2571 || (OP(p) != BRANCH
2572 && (OP(p) < BRACE_COMPLEX || OP(p) > BRACE_COMPLEX + 9)))
2573 return;
2574 regtail(OPERAND(p), val);
2575 }
2576
2577 /*
2578 * getchr() - get the next character from the pattern. We know about
2579 * magic and such, so therefore we need a lexical analyzer.
2580 */
2581
2582 /* static int curchr; */
2583 static int prevprevchr;
2584 static int prevchr;
2585 static int nextchr; /* used for ungetchr() */
2586 /*
2587 * Note: prevchr is sometimes -1 when we are not at the start,
2588 * eg in /[ ^I]^ the pattern was never found even if it existed, because ^ was
2589 * taken to be magic -- webb
2590 */
2591 static int at_start; /* True when on the first character */
2592 static int prev_at_start; /* True when on the second character */
2593
2594 static void
2595 initchr(str)
2596 char_u *str;
2597 {
2598 regparse = str;
2599 prevchr_len = 0;
2600 curchr = prevprevchr = prevchr = nextchr = -1;
2601 at_start = TRUE;
2602 prev_at_start = FALSE;
2603 }
2604
2605 static int
2606 peekchr()
2607 {
2608 static int after_slash = FALSE;
2609
2610 if (curchr == -1)
2611 {
2612 switch (curchr = regparse[0])
2613 {
2614 case '.':
2615 case '[':
2616 case '~':
2617 /* magic when 'magic' is on */
2618 if (reg_magic >= MAGIC_ON)
2619 curchr = Magic(curchr);
2620 break;
2621 case '(':
2622 case ')':
2623 case '{':
2624 case '%':
2625 case '+':
2626 case '=':
2627 case '?':
2628 case '@':
2629 case '!':
2630 case '&':
2631 case '|':
2632 case '<':
2633 case '>':
2634 case '#': /* future ext. */
2635 case '"': /* future ext. */
2636 case '\'': /* future ext. */
2637 case ',': /* future ext. */
2638 case '-': /* future ext. */
2639 case ':': /* future ext. */
2640 case ';': /* future ext. */
2641 case '`': /* future ext. */
2642 case '/': /* Can't be used in / command */
2643 /* magic only after "\v" */
2644 if (reg_magic == MAGIC_ALL)
2645 curchr = Magic(curchr);
2646 break;
2647 case '*':
2648 /* * is not magic as the very first character, eg "?*ptr", when
2649 * after '^', eg "/^*ptr" and when after "\(", "\|", "\&". But
2650 * "\(\*" is not magic, thus must be magic if "after_slash" */
2651 if (reg_magic >= MAGIC_ON
2652 && !at_start
2653 && !(prev_at_start && prevchr == Magic('^'))
2654 && (after_slash
2655 || (prevchr != Magic('(')
2656 && prevchr != Magic('&')
2657 && prevchr != Magic('|'))))
2658 curchr = Magic('*');
2659 break;
2660 case '^':
2661 /* '^' is only magic as the very first character and if it's after
2662 * "\(", "\|", "\&' or "\n" */
2663 if (reg_magic >= MAGIC_OFF
2664 && (at_start
2665 || reg_magic == MAGIC_ALL
2666 || prevchr == Magic('(')
2667 || prevchr == Magic('|')
2668 || prevchr == Magic('&')
2669 || prevchr == Magic('n')
2670 || (no_Magic(prevchr) == '('
2671 && prevprevchr == Magic('%'))))
2672 {
2673 curchr = Magic('^');
2674 at_start = TRUE;
2675 prev_at_start = FALSE;
2676 }
2677 break;
2678 case '$':
2679 /* '$' is only magic as the very last char and if it's in front of
2680 * either "\|", "\)", "\&", or "\n" */
2681 if (reg_magic >= MAGIC_OFF)
2682 {
2683 char_u *p = regparse + 1;
2684
2685 /* ignore \c \C \m and \M after '$' */
2686 while (p[0] == '\\' && (p[1] == 'c' || p[1] == 'C'
2687 || p[1] == 'm' || p[1] == 'M' || p[1] == 'Z'))
2688 p += 2;
2689 if (p[0] == NUL
2690 || (p[0] == '\\'
2691 && (p[1] == '|' || p[1] == '&' || p[1] == ')'
2692 || p[1] == 'n'))
2693 || reg_magic == MAGIC_ALL)
2694 curchr = Magic('$');
2695 }
2696 break;
2697 case '\\':
2698 {
2699 int c = regparse[1];
2700
2701 if (c == NUL)
2702 curchr = '\\'; /* trailing '\' */
2703 else if (
2704 #ifdef EBCDIC
2705 vim_strchr(META, c)
2706 #else
2707 c <= '~' && META_flags[c]
2708 #endif
2709 )
2710 {
2711 /*
2712 * META contains everything that may be magic sometimes,
2713 * except ^ and $ ("\^" and "\$" are only magic after
2714 * "\v"). We now fetch the next character and toggle its
2715 * magicness. Therefore, \ is so meta-magic that it is
2716 * not in META.
2717 */
2718 curchr = -1;
2719 prev_at_start = at_start;
2720 at_start = FALSE; /* be able to say "/\*ptr" */
2721 ++regparse;
2722 ++after_slash;
2723 peekchr();
2724 --regparse;
2725 --after_slash;
2726 curchr = toggle_Magic(curchr);
2727 }
2728 else if (vim_strchr(REGEXP_ABBR, c))
2729 {
2730 /*
2731 * Handle abbreviations, like "\t" for TAB -- webb
2732 */
2733 curchr = backslash_trans(c);
2734 }
2735 else if (reg_magic == MAGIC_NONE && (c == '$' || c == '^'))
2736 curchr = toggle_Magic(c);
2737 else
2738 {
2739 /*
2740 * Next character can never be (made) magic?
2741 * Then backslashing it won't do anything.
2742 */
2743 #ifdef FEAT_MBYTE
2744 if (has_mbyte)
2745 curchr = (*mb_ptr2char)(regparse + 1);
2746 else
2747 #endif
2748 curchr = c;
2749 }
2750 break;
2751 }
2752
2753 #ifdef FEAT_MBYTE
2754 default:
2755 if (has_mbyte)
2756 curchr = (*mb_ptr2char)(regparse);
2757 #endif
2758 }
2759 }
2760
2761 return curchr;
2762 }
2763
2764 /*
2765 * Eat one lexed character. Do this in a way that we can undo it.
2766 */
2767 static void
2768 skipchr()
2769 {
2770 /* peekchr() eats a backslash, do the same here */
2771 if (*regparse == '\\')
2772 prevchr_len = 1;
2773 else
2774 prevchr_len = 0;
2775 if (regparse[prevchr_len] != NUL)
2776 {
2777 #ifdef FEAT_MBYTE
2778 if (enc_utf8)
2779 /* exclude composing chars that mb_ptr2len does include */
2780 prevchr_len += utf_ptr2len(regparse + prevchr_len);
2781 else if (has_mbyte)
2782 prevchr_len += (*mb_ptr2len)(regparse + prevchr_len);
2783 else
2784 #endif
2785 ++prevchr_len;
2786 }
2787 regparse += prevchr_len;
2788 prev_at_start = at_start;
2789 at_start = FALSE;
2790 prevprevchr = prevchr;
2791 prevchr = curchr;
2792 curchr = nextchr; /* use previously unget char, or -1 */
2793 nextchr = -1;
2794 }
2795
2796 /*
2797 * Skip a character while keeping the value of prev_at_start for at_start.
2798 * prevchr and prevprevchr are also kept.
2799 */
2800 static void
2801 skipchr_keepstart()
2802 {
2803 int as = prev_at_start;
2804 int pr = prevchr;
2805 int prpr = prevprevchr;
2806
2807 skipchr();
2808 at_start = as;
2809 prevchr = pr;
2810 prevprevchr = prpr;
2811 }
2812
2813 static int
2814 getchr()
2815 {
2816 int chr = peekchr();
2817
2818 skipchr();
2819 return chr;
2820 }
2821
2822 /*
2823 * put character back. Works only once!
2824 */
2825 static void
2826 ungetchr()
2827 {
2828 nextchr = curchr;
2829 curchr = prevchr;
2830 prevchr = prevprevchr;
2831 at_start = prev_at_start;
2832 prev_at_start = FALSE;
2833
2834 /* Backup regparse, so that it's at the same position as before the
2835 * getchr(). */
2836 regparse -= prevchr_len;
2837 }
2838
2839 /*
2840 * Get and return the value of the hex string at the current position.
2841 * Return -1 if there is no valid hex number.
2842 * The position is updated:
2843 * blahblah\%x20asdf
2844 * before-^ ^-after
2845 * The parameter controls the maximum number of input characters. This will be
2846 * 2 when reading a \%x20 sequence and 4 when reading a \%u20AC sequence.
2847 */
2848 static int
2849 gethexchrs(maxinputlen)
2850 int maxinputlen;
2851 {
2852 int nr = 0;
2853 int c;
2854 int i;
2855
2856 for (i = 0; i < maxinputlen; ++i)
2857 {
2858 c = regparse[0];
2859 if (!vim_isxdigit(c))
2860 break;
2861 nr <<= 4;
2862 nr |= hex2nr(c);
2863 ++regparse;
2864 }
2865
2866 if (i == 0)
2867 return -1;
2868 return nr;
2869 }
2870
2871 /*
2872 * get and return the value of the decimal string immediately after the
2873 * current position. Return -1 for invalid. Consumes all digits.
2874 */
2875 static int
2876 getdecchrs()
2877 {
2878 int nr = 0;
2879 int c;
2880 int i;
2881
2882 for (i = 0; ; ++i)
2883 {
2884 c = regparse[0];
2885 if (c < '0' || c > '9')
2886 break;
2887 nr *= 10;
2888 nr += c - '0';
2889 ++regparse;
2890 }
2891
2892 if (i == 0)
2893 return -1;
2894 return nr;
2895 }
2896
2897 /*
2898 * get and return the value of the octal string immediately after the current
2899 * position. Return -1 for invalid, or 0-255 for valid. Smart enough to handle
2900 * numbers > 377 correctly (for example, 400 is treated as 40) and doesn't
2901 * treat 8 or 9 as recognised characters. Position is updated:
2902 * blahblah\%o210asdf
2903 * before-^ ^-after
2904 */
2905 static int
2906 getoctchrs()
2907 {
2908 int nr = 0;
2909 int c;
2910 int i;
2911
2912 for (i = 0; i < 3 && nr < 040; ++i)
2913 {
2914 c = regparse[0];
2915 if (c < '0' || c > '7')
2916 break;
2917 nr <<= 3;
2918 nr |= hex2nr(c);
2919 ++regparse;
2920 }
2921
2922 if (i == 0)
2923 return -1;
2924 return nr;
2925 }
2926
2927 /*
2928 * Get a number after a backslash that is inside [].
2929 * When nothing is recognized return a backslash.
2930 */
2931 static int
2932 coll_get_char()
2933 {
2934 int nr = -1;
2935
2936 switch (*regparse++)
2937 {
2938 case 'd': nr = getdecchrs(); break;
2939 case 'o': nr = getoctchrs(); break;
2940 case 'x': nr = gethexchrs(2); break;
2941 case 'u': nr = gethexchrs(4); break;
2942 case 'U': nr = gethexchrs(8); break;
2943 }
2944 if (nr < 0)
2945 {
2946 /* If getting the number fails be backwards compatible: the character
2947 * is a backslash. */
2948 --regparse;
2949 nr = '\\';
2950 }
2951 return nr;
2952 }
2953
2954 /*
2955 * read_limits - Read two integers to be taken as a minimum and maximum.
2956 * If the first character is '-', then the range is reversed.
2957 * Should end with 'end'. If minval is missing, zero is default, if maxval is
2958 * missing, a very big number is the default.
2959 */
2960 static int
2961 read_limits(minval, maxval)
2962 long *minval;
2963 long *maxval;
2964 {
2965 int reverse = FALSE;
2966 char_u *first_char;
2967 long tmp;
2968
2969 if (*regparse == '-')
2970 {
2971 /* Starts with '-', so reverse the range later */
2972 regparse++;
2973 reverse = TRUE;
2974 }
2975 first_char = regparse;
2976 *minval = getdigits(&regparse);
2977 if (*regparse == ',') /* There is a comma */
2978 {
2979 if (vim_isdigit(*++regparse))
2980 *maxval = getdigits(&regparse);
2981 else
2982 *maxval = MAX_LIMIT;
2983 }
2984 else if (VIM_ISDIGIT(*first_char))
2985 *maxval = *minval; /* It was \{n} or \{-n} */
2986 else
2987 *maxval = MAX_LIMIT; /* It was \{} or \{-} */
2988 if (*regparse == '\\')
2989 regparse++; /* Allow either \{...} or \{...\} */
2990 if (*regparse != '}')
2991 {
2992 sprintf((char *)IObuff, _("E554: Syntax error in %s{...}"),
2993 reg_magic == MAGIC_ALL ? "" : "\\");
2994 EMSG_RET_FAIL(IObuff);
2995 }
2996
2997 /*
2998 * Reverse the range if there was a '-', or make sure it is in the right
2999 * order otherwise.
3000 */
3001 if ((!reverse && *minval > *maxval) || (reverse && *minval < *maxval))
3002 {
3003 tmp = *minval;
3004 *minval = *maxval;
3005 *maxval = tmp;
3006 }
3007 skipchr(); /* let's be friends with the lexer again */
3008 return OK;
3009 }
3010
3011 /*
3012 * vim_regexec and friends
3013 */
3014
3015 /*
3016 * Global work variables for vim_regexec().
3017 */
3018
3019 /* The current match-position is remembered with these variables: */
3020 static linenr_T reglnum; /* line number, relative to first line */
3021 static char_u *regline; /* start of current line */
3022 static char_u *reginput; /* current input, points into "regline" */
3023
3024 static int need_clear_subexpr; /* subexpressions still need to be
3025 * cleared */
3026 #ifdef FEAT_SYN_HL
3027 static int need_clear_zsubexpr = FALSE; /* extmatch subexpressions
3028 * still need to be cleared */
3029 #endif
3030
3031 /*
3032 * Structure used to save the current input state, when it needs to be
3033 * restored after trying a match. Used by reg_save() and reg_restore().
3034 * Also stores the length of "backpos".
3035 */
3036 typedef struct
3037 {
3038 union
3039 {
3040 char_u *ptr; /* reginput pointer, for single-line regexp */
3041 lpos_T pos; /* reginput pos, for multi-line regexp */
3042 } rs_u;
3043 int rs_len;
3044 } regsave_T;
3045
3046 /* struct to save start/end pointer/position in for \(\) */
3047 typedef struct
3048 {
3049 union
3050 {
3051 char_u *ptr;
3052 lpos_T pos;
3053 } se_u;
3054 } save_se_T;
3055
3056 /* used for BEHIND and NOBEHIND matching */
3057 typedef struct regbehind_S
3058 {
3059 regsave_T save_after;
3060 regsave_T save_behind;
3061 int save_need_clear_subexpr;
3062 save_se_T save_start[NSUBEXP];
3063 save_se_T save_end[NSUBEXP];
3064 } regbehind_T;
3065
3066 static char_u *reg_getline __ARGS((linenr_T lnum));
3067 static long vim_regexec_both __ARGS((char_u *line, colnr_T col, proftime_T *tm));
3068 static long regtry __ARGS((regprog_T *prog, colnr_T col));
3069 static void cleanup_subexpr __ARGS((void));
3070 #ifdef FEAT_SYN_HL
3071 static void cleanup_zsubexpr __ARGS((void));
3072 #endif
3073 static void save_subexpr __ARGS((regbehind_T *bp));
3074 static void restore_subexpr __ARGS((regbehind_T *bp));
3075 static void reg_nextline __ARGS((void));
3076 static void reg_save __ARGS((regsave_T *save, garray_T *gap));
3077 static void reg_restore __ARGS((regsave_T *save, garray_T *gap));
3078 static int reg_save_equal __ARGS((regsave_T *save));
3079 static void save_se_multi __ARGS((save_se_T *savep, lpos_T *posp));
3080 static void save_se_one __ARGS((save_se_T *savep, char_u **pp));
3081
3082 /* Save the sub-expressions before attempting a match. */
3083 #define save_se(savep, posp, pp) \
3084 REG_MULTI ? save_se_multi((savep), (posp)) : save_se_one((savep), (pp))
3085
3086 /* After a failed match restore the sub-expressions. */
3087 #define restore_se(savep, posp, pp) { \
3088 if (REG_MULTI) \
3089 *(posp) = (savep)->se_u.pos; \
3090 else \
3091 *(pp) = (savep)->se_u.ptr; }
3092
3093 static int re_num_cmp __ARGS((long_u val, char_u *scan));
3094 static int regmatch __ARGS((char_u *prog));
3095 static int regrepeat __ARGS((char_u *p, long maxcount));
3096
3097 #ifdef DEBUG
3098 int regnarrate = 0;
3099 #endif
3100
3101 /*
3102 * Internal copy of 'ignorecase'. It is set at each call to vim_regexec().
3103 * Normally it gets the value of "rm_ic" or "rmm_ic", but when the pattern
3104 * contains '\c' or '\C' the value is overruled.
3105 */
3106 static int ireg_ic;
3107
3108 #ifdef FEAT_MBYTE
3109 /*
3110 * Similar to ireg_ic, but only for 'combining' characters. Set with \Z flag
3111 * in the regexp. Defaults to false, always.
3112 */
3113 static int ireg_icombine;
3114 #endif
3115
3116 /*
3117 * Copy of "rmm_maxcol": maximum column to search for a match. Zero when
3118 * there is no maximum.
3119 */
3120 static colnr_T ireg_maxcol;
3121
3122 /*
3123 * Sometimes need to save a copy of a line. Since alloc()/free() is very
3124 * slow, we keep one allocated piece of memory and only re-allocate it when
3125 * it's too small. It's freed in vim_regexec_both() when finished.
3126 */
3127 static char_u *reg_tofree = NULL;
3128 static unsigned reg_tofreelen;
3129
3130 /*
3131 * These variables are set when executing a regexp to speed up the execution.
3132 * Which ones are set depends on whether a single-line or multi-line match is
3133 * done:
3134 * single-line multi-line
3135 * reg_match &regmatch_T NULL
3136 * reg_mmatch NULL &regmmatch_T
3137 * reg_startp reg_match->startp <invalid>
3138 * reg_endp reg_match->endp <invalid>
3139 * reg_startpos <invalid> reg_mmatch->startpos
3140 * reg_endpos <invalid> reg_mmatch->endpos
3141 * reg_win NULL window in which to search
3142 * reg_buf <invalid> buffer in which to search
3143 * reg_firstlnum <invalid> first line in which to search
3144 * reg_maxline 0 last line nr
3145 * reg_line_lbr FALSE or TRUE FALSE
3146 */
3147 static regmatch_T *reg_match;
3148 static regmmatch_T *reg_mmatch;
3149 static char_u **reg_startp = NULL;
3150 static char_u **reg_endp = NULL;
3151 static lpos_T *reg_startpos = NULL;
3152 static lpos_T *reg_endpos = NULL;
3153 static win_T *reg_win;
3154 static buf_T *reg_buf;
3155 static linenr_T reg_firstlnum;
3156 static linenr_T reg_maxline;
3157 static int reg_line_lbr; /* "\n" in string is line break */
3158
3159 /* Values for rs_state in regitem_T. */
3160 typedef enum regstate_E
3161 {
3162 RS_NOPEN = 0 /* NOPEN and NCLOSE */
3163 , RS_MOPEN /* MOPEN + [0-9] */
3164 , RS_MCLOSE /* MCLOSE + [0-9] */
3165 #ifdef FEAT_SYN_HL
3166 , RS_ZOPEN /* ZOPEN + [0-9] */
3167 , RS_ZCLOSE /* ZCLOSE + [0-9] */
3168 #endif
3169 , RS_BRANCH /* BRANCH */
3170 , RS_BRCPLX_MORE /* BRACE_COMPLEX and trying one more match */
3171 , RS_BRCPLX_LONG /* BRACE_COMPLEX and trying longest match */
3172 , RS_BRCPLX_SHORT /* BRACE_COMPLEX and trying shortest match */
3173 , RS_NOMATCH /* NOMATCH */
3174 , RS_BEHIND1 /* BEHIND / NOBEHIND matching rest */
3175 , RS_BEHIND2 /* BEHIND / NOBEHIND matching behind part */
3176 , RS_STAR_LONG /* STAR/PLUS/BRACE_SIMPLE longest match */
3177 , RS_STAR_SHORT /* STAR/PLUS/BRACE_SIMPLE shortest match */
3178 } regstate_T;
3179
3180 /*
3181 * When there are alternatives a regstate_T is put on the regstack to remember
3182 * what we are doing.
3183 * Before it may be another type of item, depending on rs_state, to remember
3184 * more things.
3185 */
3186 typedef struct regitem_S
3187 {
3188 regstate_T rs_state; /* what we are doing, one of RS_ above */
3189 char_u *rs_scan; /* current node in program */
3190 union
3191 {
3192 save_se_T sesave;
3193 regsave_T regsave;
3194 } rs_un; /* room for saving reginput */
3195 short rs_no; /* submatch nr or BEHIND/NOBEHIND */
3196 } regitem_T;
3197
3198 static regitem_T *regstack_push __ARGS((regstate_T state, char_u *scan));
3199 static void regstack_pop __ARGS((char_u **scan));
3200
3201 /* used for STAR, PLUS and BRACE_SIMPLE matching */
3202 typedef struct regstar_S
3203 {
3204 int nextb; /* next byte */
3205 int nextb_ic; /* next byte reverse case */
3206 long count;
3207 long minval;
3208 long maxval;
3209 } regstar_T;
3210
3211 /* used to store input position when a BACK was encountered, so that we now if
3212 * we made any progress since the last time. */
3213 typedef struct backpos_S
3214 {
3215 char_u *bp_scan; /* "scan" where BACK was encountered */
3216 regsave_T bp_pos; /* last input position */
3217 } backpos_T;
3218
3219 /*
3220 * "regstack" and "backpos" are used by regmatch(). They are kept over calls
3221 * to avoid invoking malloc() and free() often.
3222 * "regstack" is a stack with regitem_T items, sometimes preceded by regstar_T
3223 * or regbehind_T.
3224 * "backpos_T" is a table with backpos_T for BACK
3225 */
3226 static garray_T regstack = {0, 0, 0, 0, NULL};
3227 static garray_T backpos = {0, 0, 0, 0, NULL};
3228
3229 /*
3230 * Both for regstack and backpos tables we use the following strategy of
3231 * allocation (to reduce malloc/free calls):
3232 * - Initial size is fairly small.
3233 * - When needed, the tables are grown bigger (8 times at first, double after
3234 * that).
3235 * - After executing the match we free the memory only if the array has grown.
3236 * Thus the memory is kept allocated when it's at the initial size.
3237 * This makes it fast while not keeping a lot of memory allocated.
3238 * A three times speed increase was observed when using many simple patterns.
3239 */
3240 #define REGSTACK_INITIAL 2048
3241 #define BACKPOS_INITIAL 64
3242
3243 #if defined(EXITFREE) || defined(PROTO)
3244 void
3245 free_regexp_stuff()
3246 {
3247 ga_clear(&regstack);
3248 ga_clear(&backpos);
3249 vim_free(reg_tofree);
3250 vim_free(reg_prev_sub);
3251 }
3252 #endif
3253
3254 /*
3255 * Get pointer to the line "lnum", which is relative to "reg_firstlnum".
3256 */
3257 static char_u *
3258 reg_getline(lnum)
3259 linenr_T lnum;
3260 {
3261 /* when looking behind for a match/no-match lnum is negative. But we
3262 * can't go before line 1 */
3263 if (reg_firstlnum + lnum < 1)
3264 return NULL;
3265 if (lnum > reg_maxline)
3266 /* Must have matched the "\n" in the last line. */
3267 return (char_u *)"";
3268 return ml_get_buf(reg_buf, reg_firstlnum + lnum, FALSE);
3269 }
3270
3271 static regsave_T behind_pos;
3272
3273 #ifdef FEAT_SYN_HL
3274 static char_u *reg_startzp[NSUBEXP]; /* Workspace to mark beginning */
3275 static char_u *reg_endzp[NSUBEXP]; /* and end of \z(...\) matches */
3276 static lpos_T reg_startzpos[NSUBEXP]; /* idem, beginning pos */
3277 static lpos_T reg_endzpos[NSUBEXP]; /* idem, end pos */
3278 #endif
3279
3280 /* TRUE if using multi-line regexp. */
3281 #define REG_MULTI (reg_match == NULL)
3282
3283 /*
3284 * Match a regexp against a string.
3285 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
3286 * Uses curbuf for line count and 'iskeyword'.
3287 *
3288 * Return TRUE if there is a match, FALSE if not.
3289 */
3290 int
3291 vim_regexec(rmp, line, col)
3292 regmatch_T *rmp;
3293 char_u *line; /* string to match against */
3294 colnr_T col; /* column to start looking for match */
3295 {
3296 reg_match = rmp;
3297 reg_mmatch = NULL;
3298 reg_maxline = 0;
3299 reg_line_lbr = FALSE;
3300 reg_win = NULL;
3301 ireg_ic = rmp->rm_ic;
3302 #ifdef FEAT_MBYTE
3303 ireg_icombine = FALSE;
3304 #endif
3305 ireg_maxcol = 0;
3306 return (vim_regexec_both(line, col, NULL) != 0);
3307 }
3308
3309 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) \
3310 || defined(FIND_REPLACE_DIALOG) || defined(PROTO)
3311 /*
3312 * Like vim_regexec(), but consider a "\n" in "line" to be a line break.
3313 */
3314 int
3315 vim_regexec_nl(rmp, line, col)
3316 regmatch_T *rmp;
3317 char_u *line; /* string to match against */
3318 colnr_T col; /* column to start looking for match */
3319 {
3320 reg_match = rmp;
3321 reg_mmatch = NULL;
3322 reg_maxline = 0;
3323 reg_line_lbr = TRUE;
3324 reg_win = NULL;
3325 ireg_ic = rmp->rm_ic;
3326 #ifdef FEAT_MBYTE
3327 ireg_icombine = FALSE;
3328 #endif
3329 ireg_maxcol = 0;
3330 return (vim_regexec_both(line, col, NULL) != 0);
3331 }
3332 #endif
3333
3334 /*
3335 * Match a regexp against multiple lines.
3336 * "rmp->regprog" is a compiled regexp as returned by vim_regcomp().
3337 * Uses curbuf for line count and 'iskeyword'.
3338 *
3339 * Return zero if there is no match. Return number of lines contained in the
3340 * match otherwise.
3341 */
3342 long
3343 vim_regexec_multi(rmp, win, buf, lnum, col, tm)
3344 regmmatch_T *rmp;
3345 win_T *win; /* window in which to search or NULL */
3346 buf_T *buf; /* buffer in which to search */
3347 linenr_T lnum; /* nr of line to start looking for match */
3348 colnr_T col; /* column to start looking for match */
3349 proftime_T *tm; /* timeout limit or NULL */
3350 {
3351 long r;
3352 buf_T *save_curbuf = curbuf;
3353
3354 reg_match = NULL;
3355 reg_mmatch = rmp;
3356 reg_buf = buf;
3357 reg_win = win;
3358 reg_firstlnum = lnum;
3359 reg_maxline = reg_buf->b_ml.ml_line_count - lnum;
3360 reg_line_lbr = FALSE;
3361 ireg_ic = rmp->rmm_ic;
3362 #ifdef FEAT_MBYTE
3363 ireg_icombine = FALSE;
3364 #endif
3365 ireg_maxcol = rmp->rmm_maxcol;
3366
3367 /* Need to switch to buffer "buf" to make vim_iswordc() work. */
3368 curbuf = buf;
3369 r = vim_regexec_both(NULL, col, tm);
3370 curbuf = save_curbuf;
3371
3372 return r;
3373 }
3374
3375 /*
3376 * Match a regexp against a string ("line" points to the string) or multiple
3377 * lines ("line" is NULL, use reg_getline()).
3378 */
3379 static long
3380 vim_regexec_both(line, col, tm)
3381 char_u *line;
3382 colnr_T col; /* column to start looking for match */
3383 proftime_T *tm UNUSED; /* timeout limit or NULL */
3384 {
3385 regprog_T *prog;
3386 char_u *s;
3387 long retval = 0L;
3388
3389 /* Create "regstack" and "backpos" if they are not allocated yet.
3390 * We allocate *_INITIAL amount of bytes first and then set the grow size
3391 * to much bigger value to avoid many malloc calls in case of deep regular
3392 * expressions. */
3393 if (regstack.ga_data == NULL)
3394 {
3395 /* Use an item size of 1 byte, since we push different things
3396 * onto the regstack. */
3397 ga_init2(&regstack, 1, REGSTACK_INITIAL);
3398 ga_grow(&regstack, REGSTACK_INITIAL);
3399 regstack.ga_growsize = REGSTACK_INITIAL * 8;
3400 }
3401
3402 if (backpos.ga_data == NULL)
3403 {
3404 ga_init2(&backpos, sizeof(backpos_T), BACKPOS_INITIAL);
3405 ga_grow(&backpos, BACKPOS_INITIAL);
3406 backpos.ga_growsize = BACKPOS_INITIAL * 8;
3407 }
3408
3409 if (REG_MULTI)
3410 {
3411 prog = reg_mmatch->regprog;
3412 line = reg_getline((linenr_T)0);
3413 reg_startpos = reg_mmatch->startpos;
3414 reg_endpos = reg_mmatch->endpos;
3415 }
3416 else
3417 {
3418 prog = reg_match->regprog;
3419 reg_startp = reg_match->startp;
3420 reg_endp = reg_match->endp;
3421 }
3422
3423 /* Be paranoid... */
3424 if (prog == NULL || line == NULL)
3425 {
3426 EMSG(_(e_null));
3427 goto theend;
3428 }
3429
3430 /* Check validity of program. */
3431 if (prog_magic_wrong())
3432 goto theend;
3433
3434 /* If the start column is past the maximum column: no need to try. */
3435 if (ireg_maxcol > 0 && col >= ireg_maxcol)
3436 goto theend;
3437
3438 /* If pattern contains "\c" or "\C": overrule value of ireg_ic */
3439 if (prog->regflags & RF_ICASE)
3440 ireg_ic = TRUE;
3441 else if (prog->regflags & RF_NOICASE)
3442 ireg_ic = FALSE;
3443
3444 #ifdef FEAT_MBYTE
3445 /* If pattern contains "\Z" overrule value of ireg_icombine */
3446 if (prog->regflags & RF_ICOMBINE)
3447 ireg_icombine = TRUE;
3448 #endif
3449
3450 /* If there is a "must appear" string, look for it. */
3451 if (prog->regmust != NULL)
3452 {
3453 int c;
3454
3455 #ifdef FEAT_MBYTE
3456 if (has_mbyte)
3457 c = (*mb_ptr2char)(prog->regmust);
3458 else
3459 #endif
3460 c = *prog->regmust;
3461 s = line + col;
3462
3463 /*
3464 * This is used very often, esp. for ":global". Use three versions of
3465 * the loop to avoid overhead of conditions.
3466 */
3467 if (!ireg_ic
3468 #ifdef FEAT_MBYTE
3469 && !has_mbyte
3470 #endif
3471 )
3472 while ((s = vim_strbyte(s, c)) != NULL)
3473 {
3474 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
3475 break; /* Found it. */
3476 ++s;
3477 }
3478 #ifdef FEAT_MBYTE
3479 else if (!ireg_ic || (!enc_utf8 && mb_char2len(c) > 1))
3480 while ((s = vim_strchr(s, c)) != NULL)
3481 {
3482 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
3483 break; /* Found it. */
3484 mb_ptr_adv(s);
3485 }
3486 #endif
3487 else
3488 while ((s = cstrchr(s, c)) != NULL)
3489 {
3490 if (cstrncmp(s, prog->regmust, &prog->regmlen) == 0)
3491 break; /* Found it. */
3492 mb_ptr_adv(s);
3493 }
3494 if (s == NULL) /* Not present. */
3495 goto theend;
3496 }
3497
3498 regline = line;
3499 reglnum = 0;
3500
3501 /* Simplest case: Anchored match need be tried only once. */
3502 if (prog->reganch)
3503 {
3504 int c;
3505
3506 #ifdef FEAT_MBYTE
3507 if (has_mbyte)
3508 c = (*mb_ptr2char)(regline + col);
3509 else
3510 #endif
3511 c = regline[col];
3512 if (prog->regstart == NUL
3513 || prog->regstart == c
3514 || (ireg_ic && ((
3515 #ifdef FEAT_MBYTE
3516 (enc_utf8 && utf_fold(prog->regstart) == utf_fold(c)))
3517 || (c < 255 && prog->regstart < 255 &&
3518 #endif
3519 MB_TOLOWER(prog->regstart) == MB_TOLOWER(c)))))
3520 retval = regtry(prog, col);
3521 else
3522 retval = 0;
3523 }
3524 else
3525 {
3526 #ifdef FEAT_RELTIME
3527 int tm_count = 0;
3528 #endif
3529 /* Messy cases: unanchored match. */
3530 while (!got_int)
3531 {
3532 if (prog->regstart != NUL)
3533 {
3534 /* Skip until the char we know it must start with.
3535 * Used often, do some work to avoid call overhead. */
3536 if (!ireg_ic
3537 #ifdef FEAT_MBYTE
3538 && !has_mbyte
3539 #endif
3540 )
3541 s = vim_strbyte(regline + col, prog->regstart);
3542 else
3543 s = cstrchr(regline + col, prog->regstart);
3544 if (s == NULL)
3545 {
3546 retval = 0;
3547 break;
3548 }
3549 col = (int)(s - regline);
3550 }
3551
3552 /* Check for maximum column to try. */
3553 if (ireg_maxcol > 0 && col >= ireg_maxcol)
3554 {
3555 retval = 0;
3556 break;
3557 }
3558
3559 retval = regtry(prog, col);
3560 if (retval > 0)
3561 break;
3562
3563 /* if not currently on the first line, get it again */
3564 if (reglnum != 0)
3565 {
3566 reglnum = 0;
3567 regline = reg_getline((linenr_T)0);
3568 }
3569 if (regline[col] == NUL)
3570 break;
3571 #ifdef FEAT_MBYTE
3572 if (has_mbyte)
3573 col += (*mb_ptr2len)(regline + col);
3574 else
3575 #endif
3576 ++col;
3577 #ifdef FEAT_RELTIME
3578 /* Check for timeout once in a twenty times to avoid overhead. */
3579 if (tm != NULL && ++tm_count == 20)
3580 {
3581 tm_count = 0;
3582 if (profile_passed_limit(tm))
3583 break;
3584 }
3585 #endif
3586 }
3587 }
3588
3589 theend:
3590 /* Free "reg_tofree" when it's a bit big.
3591 * Free regstack and backpos if they are bigger than their initial size. */
3592 if (reg_tofreelen > 400)
3593 {
3594 vim_free(reg_tofree);
3595 reg_tofree = NULL;
3596 }
3597 if (regstack.ga_maxlen > REGSTACK_INITIAL)
3598 ga_clear(&regstack);
3599 if (backpos.ga_maxlen > BACKPOS_INITIAL)
3600 ga_clear(&backpos);
3601
3602 return retval;
3603 }
3604
3605 #ifdef FEAT_SYN_HL
3606 static reg_extmatch_T *make_extmatch __ARGS((void));
3607
3608 /*
3609 * Create a new extmatch and mark it as referenced once.
3610 */
3611 static reg_extmatch_T *
3612 make_extmatch()
3613 {
3614 reg_extmatch_T *em;
3615
3616 em = (reg_extmatch_T *)alloc_clear((unsigned)sizeof(reg_extmatch_T));
3617 if (em != NULL)
3618 em->refcnt = 1;
3619 return em;
3620 }
3621
3622 /*
3623 * Add a reference to an extmatch.
3624 */
3625 reg_extmatch_T *
3626 ref_extmatch(em)
3627 reg_extmatch_T *em;
3628 {
3629 if (em != NULL)
3630 em->refcnt++;
3631 return em;
3632 }
3633
3634 /*
3635 * Remove a reference to an extmatch. If there are no references left, free
3636 * the info.
3637 */
3638 void
3639 unref_extmatch(em)
3640 reg_extmatch_T *em;
3641 {
3642 int i;
3643
3644 if (em != NULL && --em->refcnt <= 0)
3645 {
3646 for (i = 0; i < NSUBEXP; ++i)
3647 vim_free(em->matches[i]);
3648 vim_free(em);
3649 }
3650 }
3651 #endif
3652
3653 /*
3654 * regtry - try match of "prog" with at regline["col"].
3655 * Returns 0 for failure, number of lines contained in the match otherwise.
3656 */
3657 static long
3658 regtry(prog, col)
3659 regprog_T *prog;
3660 colnr_T col;
3661 {
3662 reginput = regline + col;
3663 need_clear_subexpr = TRUE;
3664 #ifdef FEAT_SYN_HL
3665 /* Clear the external match subpointers if necessary. */
3666 if (prog->reghasz == REX_SET)
3667 need_clear_zsubexpr = TRUE;
3668 #endif
3669
3670 if (regmatch(prog->program + 1) == 0)
3671 return 0;
3672
3673 cleanup_subexpr();
3674 if (REG_MULTI)
3675 {
3676 if (reg_startpos[0].lnum < 0)
3677 {
3678 reg_startpos[0].lnum = 0;
3679 reg_startpos[0].col = col;
3680 }
3681 if (reg_endpos[0].lnum < 0)
3682 {
3683 reg_endpos[0].lnum = reglnum;
3684 reg_endpos[0].col = (int)(reginput - regline);
3685 }
3686 else
3687 /* Use line number of "\ze". */
3688 reglnum = reg_endpos[0].lnum;
3689 }
3690 else
3691 {
3692 if (reg_startp[0] == NULL)
3693 reg_startp[0] = regline + col;
3694 if (reg_endp[0] == NULL)
3695 reg_endp[0] = reginput;
3696 }
3697 #ifdef FEAT_SYN_HL
3698 /* Package any found \z(...\) matches for export. Default is none. */
3699 unref_extmatch(re_extmatch_out);
3700 re_extmatch_out = NULL;
3701
3702 if (prog->reghasz == REX_SET)
3703 {
3704 int i;
3705
3706 cleanup_zsubexpr();
3707 re_extmatch_out = make_extmatch();
3708 for (i = 0; i < NSUBEXP; i++)
3709 {
3710 if (REG_MULTI)
3711 {
3712 /* Only accept single line matches. */
3713 if (reg_startzpos[i].lnum >= 0
3714 && reg_endzpos[i].lnum == reg_startzpos[i].lnum)
3715 re_extmatch_out->matches[i] =
3716 vim_strnsave(reg_getline(reg_startzpos[i].lnum)
3717 + reg_startzpos[i].col,
3718 reg_endzpos[i].col - reg_startzpos[i].col);
3719 }
3720 else
3721 {
3722 if (reg_startzp[i] != NULL && reg_endzp[i] != NULL)
3723 re_extmatch_out->matches[i] =
3724 vim_strnsave(reg_startzp[i],
3725 (int)(reg_endzp[i] - reg_startzp[i]));
3726 }
3727 }
3728 }
3729 #endif
3730 return 1 + reglnum;
3731 }
3732
3733 #ifdef FEAT_MBYTE
3734 static int reg_prev_class __ARGS((void));
3735
3736 /*
3737 * Get class of previous character.
3738 */
3739 static int
3740 reg_prev_class()
3741 {
3742 if (reginput > regline)
3743 return mb_get_class(reginput - 1
3744 - (*mb_head_off)(regline, reginput - 1));
3745 return -1;
3746 }
3747
3748 #endif
3749 #define ADVANCE_REGINPUT() mb_ptr_adv(reginput)
3750
3751 /*
3752 * The arguments from BRACE_LIMITS are stored here. They are actually local
3753 * to regmatch(), but they are here to reduce the amount of stack space used
3754 * (it can be called recursively many times).
3755 */
3756 static long bl_minval;
3757 static long bl_maxval;
3758
3759 /*
3760 * regmatch - main matching routine
3761 *
3762 * Conceptually the strategy is simple: Check to see whether the current node
3763 * matches, push an item onto the regstack and loop to see whether the rest
3764 * matches, and then act accordingly. In practice we make some effort to
3765 * avoid using the regstack, in particular by going through "ordinary" nodes
3766 * (that don't need to know whether the rest of the match failed) by a nested
3767 * loop.
3768 *
3769 * Returns TRUE when there is a match. Leaves reginput and reglnum just after
3770 * the last matched character.
3771 * Returns FALSE when there is no match. Leaves reginput and reglnum in an
3772 * undefined state!
3773 */
3774 static int
3775 regmatch(scan)
3776 char_u *scan; /* Current node. */
3777 {
3778 char_u *next; /* Next node. */
3779 int op;
3780 int c;
3781 regitem_T *rp;
3782 int no;
3783 int status; /* one of the RA_ values: */
3784 #define RA_FAIL 1 /* something failed, abort */
3785 #define RA_CONT 2 /* continue in inner loop */
3786 #define RA_BREAK 3 /* break inner loop */
3787 #define RA_MATCH 4 /* successful match */
3788 #define RA_NOMATCH 5 /* didn't match */
3789
3790 /* Make "regstack" and "backpos" empty. They are allocated and freed in
3791 * vim_regexec_both() to reduce malloc()/free() calls. */
3792 regstack.ga_len = 0;
3793 backpos.ga_len = 0;
3794
3795 /*
3796 * Repeat until "regstack" is empty.
3797 */
3798 for (;;)
3799 {
3800 /* Some patterns my cause a long time to match, even though they are not
3801 * illegal. E.g., "\([a-z]\+\)\+Q". Allow breaking them with CTRL-C. */
3802 fast_breakcheck();
3803
3804 #ifdef DEBUG
3805 if (scan != NULL && regnarrate)
3806 {
3807 mch_errmsg(regprop(scan));
3808 mch_errmsg("(\n");
3809 }
3810 #endif
3811
3812 /*
3813 * Repeat for items that can be matched sequentially, without using the
3814 * regstack.
3815 */
3816 for (;;)
3817 {
3818 if (got_int || scan == NULL)
3819 {
3820 status = RA_FAIL;
3821 break;
3822 }
3823 status = RA_CONT;
3824
3825 #ifdef DEBUG
3826 if (regnarrate)
3827 {
3828 mch_errmsg(regprop(scan));
3829 mch_errmsg("...\n");
3830 # ifdef FEAT_SYN_HL
3831 if (re_extmatch_in != NULL)
3832 {
3833 int i;
3834
3835 mch_errmsg(_("External submatches:\n"));
3836 for (i = 0; i < NSUBEXP; i++)
3837 {
3838 mch_errmsg(" \"");
3839 if (re_extmatch_in->matches[i] != NULL)
3840 mch_errmsg(re_extmatch_in->matches[i]);
3841 mch_errmsg("\"\n");
3842 }
3843 }
3844 # endif
3845 }
3846 #endif
3847 next = regnext(scan);
3848
3849 op = OP(scan);
3850 /* Check for character class with NL added. */
3851 if (!reg_line_lbr && WITH_NL(op) && REG_MULTI
3852 && *reginput == NUL && reglnum <= reg_maxline)
3853 {
3854 reg_nextline();
3855 }
3856 else if (reg_line_lbr && WITH_NL(op) && *reginput == '\n')
3857 {
3858 ADVANCE_REGINPUT();
3859 }
3860 else
3861 {
3862 if (WITH_NL(op))
3863 op -= ADD_NL;
3864 #ifdef FEAT_MBYTE
3865 if (has_mbyte)
3866 c = (*mb_ptr2char)(reginput);
3867 else
3868 #endif
3869 c = *reginput;
3870 switch (op)
3871 {
3872 case BOL:
3873 if (reginput != regline)
3874 status = RA_NOMATCH;
3875 break;
3876
3877 case EOL:
3878 if (c != NUL)
3879 status = RA_NOMATCH;
3880 break;
3881
3882 case RE_BOF:
3883 /* We're not at the beginning of the file when below the first
3884 * line where we started, not at the start of the line or we
3885 * didn't start at the first line of the buffer. */
3886 if (reglnum != 0 || reginput != regline
3887 || (REG_MULTI && reg_firstlnum > 1))
3888 status = RA_NOMATCH;
3889 break;
3890
3891 case RE_EOF:
3892 if (reglnum != reg_maxline || c != NUL)
3893 status = RA_NOMATCH;
3894 break;
3895
3896 case CURSOR:
3897 /* Check if the buffer is in a window and compare the
3898 * reg_win->w_cursor position to the match position. */
3899 if (reg_win == NULL
3900 || (reglnum + reg_firstlnum != reg_win->w_cursor.lnum)
3901 || ((colnr_T)(reginput - regline) != reg_win->w_cursor.col))
3902 status = RA_NOMATCH;
3903 break;
3904
3905 case RE_MARK:
3906 /* Compare the mark position to the match position. NOTE: Always
3907 * uses the current buffer. */
3908 {
3909 int mark = OPERAND(scan)[0];
3910 int cmp = OPERAND(scan)[1];
3911 pos_T *pos;
3912
3913 pos = getmark(mark, FALSE);
3914 if (pos == NULL /* mark doesn't exist */
3915 || pos->lnum <= 0 /* mark isn't set (in curbuf) */
3916 || (pos->lnum == reglnum + reg_firstlnum
3917 ? (pos->col == (colnr_T)(reginput - regline)
3918 ? (cmp == '<' || cmp == '>')
3919 : (pos->col < (colnr_T)(reginput - regline)
3920 ? cmp != '>'
3921 : cmp != '<'))
3922 : (pos->lnum < reglnum + reg_firstlnum
3923 ? cmp != '>'
3924 : cmp != '<')))
3925 status = RA_NOMATCH;
3926 }
3927 break;
3928
3929 case RE_VISUAL:
3930 #ifdef FEAT_VISUAL
3931 /* Check if the buffer is the current buffer. and whether the
3932 * position is inside the Visual area. */
3933 if (reg_buf != curbuf || VIsual.lnum == 0)
3934 status = RA_NOMATCH;
3935 else
3936 {
3937 pos_T top, bot;
3938 linenr_T lnum;
3939 colnr_T col;
3940 win_T *wp = reg_win == NULL ? curwin : reg_win;
3941 int mode;
3942
3943 if (VIsual_active)
3944 {
3945 if (lt(VIsual, wp->w_cursor))
3946 {
3947 top = VIsual;
3948 bot = wp->w_cursor;
3949 }
3950 else
3951 {
3952 top = wp->w_cursor;
3953 bot = VIsual;
3954 }
3955 mode = VIsual_mode;
3956 }
3957 else
3958 {
3959 if (lt(curbuf->b_visual.vi_start, curbuf->b_visual.vi_end))
3960 {
3961 top = curbuf->b_visual.vi_start;
3962 bot = curbuf->b_visual.vi_end;
3963 }
3964 else
3965 {
3966 top = curbuf->b_visual.vi_end;
3967 bot = curbuf->b_visual.vi_start;
3968 }
3969 mode = curbuf->b_visual.vi_mode;
3970 }
3971 lnum = reglnum + reg_firstlnum;
3972 col = (colnr_T)(reginput - regline);
3973 if (lnum < top.lnum || lnum > bot.lnum)
3974 status = RA_NOMATCH;
3975 else if (mode == 'v')
3976 {
3977 if ((lnum == top.lnum && col < top.col)
3978 || (lnum == bot.lnum
3979 && col >= bot.col + (*p_sel != 'e')))
3980 status = RA_NOMATCH;
3981 }
3982 else if (mode == Ctrl_V)
3983 {
3984 colnr_T start, end;
3985 colnr_T start2, end2;
3986 colnr_T cols;
3987
3988 getvvcol(wp, &top, &start, NULL, &end);
3989 getvvcol(wp, &bot, &start2, NULL, &end2);
3990 if (start2 < start)
3991 start = start2;
3992 if (end2 > end)
3993 end = end2;
3994 if (top.col == MAXCOL || bot.col == MAXCOL)
3995 end = MAXCOL;
3996 cols = win_linetabsize(wp,
3997 regline, (colnr_T)(reginput - regline));
3998 if (cols < start || cols > end - (*p_sel == 'e'))
3999 status = RA_NOMATCH;
4000 }
4001 }
4002 #else
4003 status = RA_NOMATCH;
4004 #endif
4005 break;
4006
4007 case RE_LNUM:
4008 if (!REG_MULTI || !re_num_cmp((long_u)(reglnum + reg_firstlnum),
4009 scan))
4010 status = RA_NOMATCH;
4011 break;
4012
4013 case RE_COL:
4014 if (!re_num_cmp((long_u)(reginput - regline) + 1, scan))
4015 status = RA_NOMATCH;
4016 break;
4017
4018 case RE_VCOL:
4019 if (!re_num_cmp((long_u)win_linetabsize(
4020 reg_win == NULL ? curwin : reg_win,
4021 regline, (colnr_T)(reginput - regline)) + 1, scan))
4022 status = RA_NOMATCH;
4023 break;
4024
4025 case BOW: /* \<word; reginput points to w */
4026 if (c == NUL) /* Can't match at end of line */
4027 status = RA_NOMATCH;
4028 #ifdef FEAT_MBYTE
4029 else if (has_mbyte)
4030 {
4031 int this_class;
4032
4033 /* Get class of current and previous char (if it exists). */
4034 this_class = mb_get_class(reginput);
4035 if (this_class <= 1)
4036 status = RA_NOMATCH; /* not on a word at all */
4037 else if (reg_prev_class() == this_class)
4038 status = RA_NOMATCH; /* previous char is in same word */
4039 }
4040 #endif
4041 else
4042 {
4043 if (!vim_iswordc(c)
4044 || (reginput > regline && vim_iswordc(reginput[-1])))
4045 status = RA_NOMATCH;
4046 }
4047 break;
4048
4049 case EOW: /* word\>; reginput points after d */
4050 if (reginput == regline) /* Can't match at start of line */
4051 status = RA_NOMATCH;
4052 #ifdef FEAT_MBYTE
4053 else if (has_mbyte)
4054 {
4055 int this_class, prev_class;
4056
4057 /* Get class of current and previous char (if it exists). */
4058 this_class = mb_get_class(reginput);
4059 prev_class = reg_prev_class();
4060 if (this_class == prev_class
4061 || prev_class == 0 || prev_class == 1)
4062 status = RA_NOMATCH;
4063 }
4064 #endif
4065 else
4066 {
4067 if (!vim_iswordc(reginput[-1])
4068 || (reginput[0] != NUL && vim_iswordc(c)))
4069 status = RA_NOMATCH;
4070 }
4071 break; /* Matched with EOW */
4072
4073 case ANY:
4074 if (c == NUL)
4075 status = RA_NOMATCH;
4076 else
4077 ADVANCE_REGINPUT();
4078 break;
4079
4080 case IDENT:
4081 if (!vim_isIDc(c))
4082 status = RA_NOMATCH;
4083 else
4084 ADVANCE_REGINPUT();
4085 break;
4086
4087 case SIDENT:
4088 if (VIM_ISDIGIT(*reginput) || !vim_isIDc(c))
4089 status = RA_NOMATCH;
4090 else
4091 ADVANCE_REGINPUT();
4092 break;
4093
4094 case KWORD:
4095 if (!vim_iswordp(reginput))
4096 status = RA_NOMATCH;
4097 else
4098 ADVANCE_REGINPUT();
4099 break;
4100
4101 case SKWORD:
4102 if (VIM_ISDIGIT(*reginput) || !vim_iswordp(reginput))
4103 status = RA_NOMATCH;
4104 else
4105 ADVANCE_REGINPUT();
4106 break;
4107
4108 case FNAME:
4109 if (!vim_isfilec(c))
4110 status = RA_NOMATCH;
4111 else
4112 ADVANCE_REGINPUT();
4113 break;
4114
4115 case SFNAME:
4116 if (VIM_ISDIGIT(*reginput) || !vim_isfilec(c))
4117 status = RA_NOMATCH;
4118 else
4119 ADVANCE_REGINPUT();
4120 break;
4121
4122 case PRINT:
4123 if (ptr2cells(reginput) != 1)
4124 status = RA_NOMATCH;
4125 else
4126 ADVANCE_REGINPUT();
4127 break;
4128
4129 case SPRINT:
4130 if (VIM_ISDIGIT(*reginput) || ptr2cells(reginput) != 1)
4131 status = RA_NOMATCH;
4132 else
4133 ADVANCE_REGINPUT();
4134 break;
4135
4136 case WHITE:
4137 if (!vim_iswhite(c))
4138 status = RA_NOMATCH;
4139 else
4140 ADVANCE_REGINPUT();
4141 break;
4142
4143 case NWHITE:
4144 if (c == NUL || vim_iswhite(c))
4145 status = RA_NOMATCH;
4146 else
4147 ADVANCE_REGINPUT();
4148 break;
4149
4150 case DIGIT:
4151 if (!ri_digit(c))
4152 status = RA_NOMATCH;
4153 else
4154 ADVANCE_REGINPUT();
4155 break;
4156
4157 case NDIGIT:
4158 if (c == NUL || ri_digit(c))
4159 status = RA_NOMATCH;
4160 else
4161 ADVANCE_REGINPUT();
4162 break;
4163
4164 case HEX:
4165 if (!ri_hex(c))
4166 status = RA_NOMATCH;
4167 else
4168 ADVANCE_REGINPUT();
4169 break;
4170
4171 case NHEX:
4172 if (c == NUL || ri_hex(c))
4173 status = RA_NOMATCH;
4174 else
4175 ADVANCE_REGINPUT();
4176 break;
4177
4178 case OCTAL:
4179 if (!ri_octal(c))
4180 status = RA_NOMATCH;
4181 else
4182 ADVANCE_REGINPUT();
4183 break;
4184
4185 case NOCTAL:
4186 if (c == NUL || ri_octal(c))
4187 status = RA_NOMATCH;
4188 else
4189 ADVANCE_REGINPUT();
4190 break;
4191
4192 case WORD:
4193 if (!ri_word(c))
4194 status = RA_NOMATCH;
4195 else
4196 ADVANCE_REGINPUT();
4197 break;
4198
4199 case NWORD:
4200 if (c == NUL || ri_word(c))
4201 status = RA_NOMATCH;
4202 else
4203 ADVANCE_REGINPUT();
4204 break;
4205
4206 case HEAD:
4207 if (!ri_head(c))
4208 status = RA_NOMATCH;
4209 else
4210 ADVANCE_REGINPUT();
4211 break;
4212
4213 case NHEAD:
4214 if (c == NUL || ri_head(c))
4215 status = RA_NOMATCH;
4216 else
4217 ADVANCE_REGINPUT();
4218 break;
4219
4220 case ALPHA:
4221 if (!ri_alpha(c))
4222 status = RA_NOMATCH;
4223 else
4224 ADVANCE_REGINPUT();
4225 break;
4226
4227 case NALPHA:
4228 if (c == NUL || ri_alpha(c))
4229 status = RA_NOMATCH;
4230 else
4231 ADVANCE_REGINPUT();
4232 break;
4233
4234 case LOWER:
4235 if (!ri_lower(c))
4236 status = RA_NOMATCH;
4237 else
4238 ADVANCE_REGINPUT();
4239 break;
4240
4241 case NLOWER:
4242 if (c == NUL || ri_lower(c))
4243 status = RA_NOMATCH;
4244 else
4245 ADVANCE_REGINPUT();
4246 break;
4247
4248 case UPPER:
4249 if (!ri_upper(c))
4250 status = RA_NOMATCH;
4251 else
4252 ADVANCE_REGINPUT();
4253 break;
4254
4255 case NUPPER:
4256 if (c == NUL || ri_upper(c))
4257 status = RA_NOMATCH;
4258 else
4259 ADVANCE_REGINPUT();
4260 break;
4261
4262 case EXACTLY:
4263 {
4264 int len;
4265 char_u *opnd;
4266
4267 opnd = OPERAND(scan);
4268 /* Inline the first byte, for speed. */
4269 if (*opnd != *reginput
4270 && (!ireg_ic || (
4271 #ifdef FEAT_MBYTE
4272 !enc_utf8 &&
4273 #endif
4274 MB_TOLOWER(*opnd) != MB_TOLOWER(*reginput))))
4275 status = RA_NOMATCH;
4276 else if (*opnd == NUL)
4277 {
4278 /* match empty string always works; happens when "~" is
4279 * empty. */
4280 }
4281 else if (opnd[1] == NUL
4282 #ifdef FEAT_MBYTE
4283 && !(enc_utf8 && ireg_ic)
4284 #endif
4285 )
4286 ++reginput; /* matched a single char */
4287 else
4288 {
4289 len = (int)STRLEN(opnd);
4290 /* Need to match first byte again for multi-byte. */
4291 if (cstrncmp(opnd, reginput, &len) != 0)
4292 status = RA_NOMATCH;
4293 #ifdef FEAT_MBYTE
4294 /* Check for following composing character. */
4295 else if (enc_utf8
4296 && UTF_COMPOSINGLIKE(reginput, reginput + len))
4297 {
4298 /* raaron: This code makes a composing character get
4299 * ignored, which is the correct behavior (sometimes)
4300 * for voweled Hebrew texts. */
4301 if (!ireg_icombine)
4302 status = RA_NOMATCH;
4303 }
4304 #endif
4305 else
4306 reginput += len;
4307 }
4308 }
4309 break;
4310
4311 case ANYOF:
4312 case ANYBUT:
4313 if (c == NUL)
4314 status = RA_NOMATCH;
4315 else if ((cstrchr(OPERAND(scan), c) == NULL) == (op == ANYOF))
4316 status = RA_NOMATCH;
4317 else
4318 ADVANCE_REGINPUT();
4319 break;
4320
4321 #ifdef FEAT_MBYTE
4322 case MULTIBYTECODE:
4323 if (has_mbyte)
4324 {
4325 int i, len;
4326 char_u *opnd;
4327 int opndc = 0, inpc;
4328
4329 opnd = OPERAND(scan);
4330 /* Safety check (just in case 'encoding' was changed since
4331 * compiling the program). */
4332 if ((len = (*mb_ptr2len)(opnd)) < 2)
4333 {
4334 status = RA_NOMATCH;
4335 break;
4336 }
4337 if (enc_utf8)
4338 opndc = mb_ptr2char(opnd);
4339 if (enc_utf8 && utf_iscomposing(opndc))
4340 {
4341 /* When only a composing char is given match at any
4342 * position where that composing char appears. */
4343 status = RA_NOMATCH;
4344 for (i = 0; reginput[i] != NUL; i += utf_char2len(inpc))
4345 {
4346 inpc = mb_ptr2char(reginput + i);
4347 if (!utf_iscomposing(inpc))
4348 {
4349 if (i > 0)
4350 break;
4351 }
4352 else if (opndc == inpc)
4353 {
4354 /* Include all following composing chars. */
4355 len = i + mb_ptr2len(reginput + i);
4356 status = RA_MATCH;
4357 break;
4358 }
4359 }
4360 }
4361 else
4362 for (i = 0; i < len; ++i)
4363 if (opnd[i] != reginput[i])
4364 {
4365 status = RA_NOMATCH;
4366 break;
4367 }
4368 reginput += len;
4369 }
4370 else
4371 status = RA_NOMATCH;
4372 break;
4373 #endif
4374
4375 case NOTHING:
4376 break;
4377
4378 case BACK:
4379 {
4380 int i;
4381 backpos_T *bp;
4382
4383 /*
4384 * When we run into BACK we need to check if we don't keep
4385 * looping without matching any input. The second and later
4386 * times a BACK is encountered it fails if the input is still
4387 * at the same position as the previous time.
4388 * The positions are stored in "backpos" and found by the
4389 * current value of "scan", the position in the RE program.
4390 */
4391 bp = (backpos_T *)backpos.ga_data;
4392 for (i = 0; i < backpos.ga_len; ++i)
4393 if (bp[i].bp_scan == scan)
4394 break;
4395 if (i == backpos.ga_len)
4396 {
4397 /* First time at this BACK, make room to store the pos. */
4398 if (ga_grow(&backpos, 1) == FAIL)
4399 status = RA_FAIL;
4400 else
4401 {
4402 /* get "ga_data" again, it may have changed */
4403 bp = (backpos_T *)backpos.ga_data;
4404 bp[i].bp_scan = scan;
4405 ++backpos.ga_len;
4406 }
4407 }
4408 else if (reg_save_equal(&bp[i].bp_pos))
4409 /* Still at same position as last time, fail. */
4410 status = RA_NOMATCH;
4411
4412 if (status != RA_FAIL && status != RA_NOMATCH)
4413 reg_save(&bp[i].bp_pos, &backpos);
4414 }
4415 break;
4416
4417 case MOPEN + 0: /* Match start: \zs */
4418 case MOPEN + 1: /* \( */
4419 case MOPEN + 2:
4420 case MOPEN + 3:
4421 case MOPEN + 4:
4422 case MOPEN + 5:
4423 case MOPEN + 6:
4424 case MOPEN + 7:
4425 case MOPEN + 8:
4426 case MOPEN + 9:
4427 {
4428 no = op - MOPEN;
4429 cleanup_subexpr();
4430 rp = regstack_push(RS_MOPEN, scan);
4431 if (rp == NULL)
4432 status = RA_FAIL;
4433 else
4434 {
4435 rp->rs_no = no;
4436 save_se(&rp->rs_un.sesave, &reg_startpos[no],
4437 &reg_startp[no]);
4438 /* We simply continue and handle the result when done. */
4439 }
4440 }
4441 break;
4442
4443 case NOPEN: /* \%( */
4444 case NCLOSE: /* \) after \%( */
4445 if (regstack_push(RS_NOPEN, scan) == NULL)
4446 status = RA_FAIL;
4447 /* We simply continue and handle the result when done. */
4448 break;
4449
4450 #ifdef FEAT_SYN_HL
4451 case ZOPEN + 1:
4452 case ZOPEN + 2:
4453 case ZOPEN + 3:
4454 case ZOPEN + 4:
4455 case ZOPEN + 5:
4456 case ZOPEN + 6:
4457 case ZOPEN + 7:
4458 case ZOPEN + 8:
4459 case ZOPEN + 9:
4460 {
4461 no = op - ZOPEN;
4462 cleanup_zsubexpr();
4463 rp = regstack_push(RS_ZOPEN, scan);
4464 if (rp == NULL)
4465 status = RA_FAIL;
4466 else
4467 {
4468 rp->rs_no = no;
4469 save_se(&rp->rs_un.sesave, &reg_startzpos[no],
4470 &reg_startzp[no]);
4471 /* We simply continue and handle the result when done. */
4472 }
4473 }
4474 break;
4475 #endif
4476
4477 case MCLOSE + 0: /* Match end: \ze */
4478 case MCLOSE + 1: /* \) */
4479 case MCLOSE + 2:
4480 case MCLOSE + 3:
4481 case MCLOSE + 4:
4482 case MCLOSE + 5:
4483 case MCLOSE + 6:
4484 case MCLOSE + 7:
4485 case MCLOSE + 8:
4486 case MCLOSE + 9:
4487 {
4488 no = op - MCLOSE;
4489 cleanup_subexpr();
4490 rp = regstack_push(RS_MCLOSE, scan);
4491 if (rp == NULL)
4492 status = RA_FAIL;
4493 else
4494 {
4495 rp->rs_no = no;
4496 save_se(&rp->rs_un.sesave, &reg_endpos[no], &reg_endp[no]);
4497 /* We simply continue and handle the result when done. */
4498 }
4499 }
4500 break;
4501
4502 #ifdef FEAT_SYN_HL
4503 case ZCLOSE + 1: /* \) after \z( */
4504 case ZCLOSE + 2:
4505 case ZCLOSE + 3:
4506 case ZCLOSE + 4:
4507 case ZCLOSE + 5:
4508 case ZCLOSE + 6:
4509 case ZCLOSE + 7:
4510 case ZCLOSE + 8:
4511 case ZCLOSE + 9:
4512 {
4513 no = op - ZCLOSE;
4514 cleanup_zsubexpr();
4515 rp = regstack_push(RS_ZCLOSE, scan);
4516 if (rp == NULL)
4517 status = RA_FAIL;
4518 else
4519 {
4520 rp->rs_no = no;
4521 save_se(&rp->rs_un.sesave, &reg_endzpos[no],
4522 &reg_endzp[no]);
4523 /* We simply continue and handle the result when done. */
4524 }
4525 }
4526 break;
4527 #endif
4528
4529 case BACKREF + 1:
4530 case BACKREF + 2:
4531 case BACKREF + 3:
4532 case BACKREF + 4:
4533 case BACKREF + 5:
4534 case BACKREF + 6:
4535 case BACKREF + 7:
4536 case BACKREF + 8:
4537 case BACKREF + 9:
4538 {
4539 int len;
4540 linenr_T clnum;
4541 colnr_T ccol;
4542 char_u *p;
4543
4544 no = op - BACKREF;
4545 cleanup_subexpr();
4546 if (!REG_MULTI) /* Single-line regexp */
4547 {
4548 if (reg_startp[no] == NULL || reg_endp[no] == NULL)
4549 {
4550 /* Backref was not set: Match an empty string. */
4551 len = 0;
4552 }
4553 else
4554 {
4555 /* Compare current input with back-ref in the same
4556 * line. */
4557 len = (int)(reg_endp[no] - reg_startp[no]);
4558 if (cstrncmp(reg_startp[no], reginput, &len) != 0)
4559 status = RA_NOMATCH;
4560 }
4561 }
4562 else /* Multi-line regexp */
4563 {
4564 if (reg_startpos[no].lnum < 0 || reg_endpos[no].lnum < 0)
4565 {
4566 /* Backref was not set: Match an empty string. */
4567 len = 0;
4568 }
4569 else
4570 {
4571 if (reg_startpos[no].lnum == reglnum
4572 && reg_endpos[no].lnum == reglnum)
4573 {
4574 /* Compare back-ref within the current line. */
4575 len = reg_endpos[no].col - reg_startpos[no].col;
4576 if (cstrncmp(regline + reg_startpos[no].col,
4577 reginput, &len) != 0)
4578 status = RA_NOMATCH;
4579 }
4580 else
4581 {
4582 /* Messy situation: Need to compare between two
4583 * lines. */
4584 ccol = reg_startpos[no].col;
4585 clnum = reg_startpos[no].lnum;
4586 for (;;)
4587 {
4588 /* Since getting one line may invalidate
4589 * the other, need to make copy. Slow! */
4590 if (regline != reg_tofree)
4591 {
4592 len = (int)STRLEN(regline);
4593 if (reg_tofree == NULL
4594 || len >= (int)reg_tofreelen)
4595 {
4596 len += 50; /* get some extra */
4597 vim_free(reg_tofree);
4598 reg_tofree = alloc(len);
4599 if (reg_tofree == NULL)
4600 {
4601 status = RA_FAIL; /* outof memory!*/
4602 break;
4603 }
4604 reg_tofreelen = len;
4605 }
4606 STRCPY(reg_tofree, regline);
4607 reginput = reg_tofree
4608 + (reginput - regline);
4609 regline = reg_tofree;
4610 }
4611
4612 /* Get the line to compare with. */
4613 p = reg_getline(clnum);
4614 if (clnum == reg_endpos[no].lnum)
4615 len = reg_endpos[no].col - ccol;
4616 else
4617 len = (int)STRLEN(p + ccol);
4618
4619 if (cstrncmp(p + ccol, reginput, &len) != 0)
4620 {
4621 status = RA_NOMATCH; /* doesn't match */
4622 break;
4623 }
4624 if (clnum == reg_endpos[no].lnum)
4625 break; /* match and at end! */
4626 if (reglnum >= reg_maxline)
4627 {
4628 status = RA_NOMATCH; /* text too short */
4629 break;
4630 }
4631
4632 /* Advance to next line. */
4633 reg_nextline();
4634 ++clnum;
4635 ccol = 0;
4636 if (got_int)
4637 {
4638 status = RA_FAIL;
4639 break;
4640 }
4641 }
4642
4643 /* found a match! Note that regline may now point
4644 * to a copy of the line, that should not matter. */
4645 }
4646 }
4647 }
4648
4649 /* Matched the backref, skip over it. */
4650 reginput += len;
4651 }
4652 break;
4653
4654 #ifdef FEAT_SYN_HL
4655 case ZREF + 1:
4656 case ZREF + 2:
4657 case ZREF + 3:
4658 case ZREF + 4:
4659 case ZREF + 5:
4660 case ZREF + 6:
4661 case ZREF + 7:
4662 case ZREF + 8:
4663 case ZREF + 9:
4664 {
4665 int len;
4666
4667 cleanup_zsubexpr();
4668 no = op - ZREF;
4669 if (re_extmatch_in != NULL
4670 && re_extmatch_in->matches[no] != NULL)
4671 {
4672 len = (int)STRLEN(re_extmatch_in->matches[no]);
4673 if (cstrncmp(re_extmatch_in->matches[no],
4674 reginput, &len) != 0)
4675 status = RA_NOMATCH;
4676 else
4677 reginput += len;
4678 }
4679 else
4680 {
4681 /* Backref was not set: Match an empty string. */
4682 }
4683 }
4684 break;
4685 #endif
4686
4687 case BRANCH:
4688 {
4689 if (OP(next) != BRANCH) /* No choice. */
4690 next = OPERAND(scan); /* Avoid recursion. */
4691 else
4692 {
4693 rp = regstack_push(RS_BRANCH, scan);
4694 if (rp == NULL)
4695 status = RA_FAIL;
4696 else
4697 status = RA_BREAK; /* rest is below */
4698 }
4699 }
4700 break;
4701
4702 case BRACE_LIMITS:
4703 {
4704 if (OP(next) == BRACE_SIMPLE)
4705 {
4706 bl_minval = OPERAND_MIN(scan);
4707 bl_maxval = OPERAND_MAX(scan);
4708 }
4709 else if (OP(next) >= BRACE_COMPLEX
4710 && OP(next) < BRACE_COMPLEX + 10)
4711 {
4712 no = OP(next) - BRACE_COMPLEX;
4713 brace_min[no] = OPERAND_MIN(scan);
4714 brace_max[no] = OPERAND_MAX(scan);
4715 brace_count[no] = 0;
4716 }
4717 else
4718 {
4719 EMSG(_(e_internal)); /* Shouldn't happen */
4720 status = RA_FAIL;
4721 }
4722 }
4723 break;
4724
4725 case BRACE_COMPLEX + 0:
4726 case BRACE_COMPLEX + 1:
4727 case BRACE_COMPLEX + 2:
4728 case BRACE_COMPLEX + 3:
4729 case BRACE_COMPLEX + 4:
4730 case BRACE_COMPLEX + 5:
4731 case BRACE_COMPLEX + 6:
4732 case BRACE_COMPLEX + 7:
4733 case BRACE_COMPLEX + 8:
4734 case BRACE_COMPLEX + 9:
4735 {
4736 no = op - BRACE_COMPLEX;
4737 ++brace_count[no];
4738
4739 /* If not matched enough times yet, try one more */
4740 if (brace_count[no] <= (brace_min[no] <= brace_max[no]
4741 ? brace_min[no] : brace_max[no]))
4742 {
4743 rp = regstack_push(RS_BRCPLX_MORE, scan);
4744 if (rp == NULL)
4745 status = RA_FAIL;
4746 else
4747 {
4748 rp->rs_no = no;
4749 reg_save(&rp->rs_un.regsave, &backpos);
4750 next = OPERAND(scan);
4751 /* We continue and handle the result when done. */
4752 }
4753 break;
4754 }
4755
4756 /* If matched enough times, may try matching some more */
4757 if (brace_min[no] <= brace_max[no])
4758 {
4759 /* Range is the normal way around, use longest match */
4760 if (brace_count[no] <= brace_max[no])
4761 {
4762 rp = regstack_push(RS_BRCPLX_LONG, scan);
4763 if (rp == NULL)
4764 status = RA_FAIL;
4765 else
4766 {
4767 rp->rs_no = no;
4768 reg_save(&rp->rs_un.regsave, &backpos);
4769 next = OPERAND(scan);
4770 /* We continue and handle the result when done. */
4771 }
4772 }
4773 }
4774 else
4775 {
4776 /* Range is backwards, use shortest match first */
4777 if (brace_count[no] <= brace_min[no])
4778 {
4779 rp = regstack_push(RS_BRCPLX_SHORT, scan);
4780 if (rp == NULL)
4781 status = RA_FAIL;
4782 else
4783 {
4784 reg_save(&rp->rs_un.regsave, &backpos);
4785 /* We continue and handle the result when done. */
4786 }
4787 }
4788 }
4789 }
4790 break;
4791
4792 case BRACE_SIMPLE:
4793 case STAR:
4794 case PLUS:
4795 {
4796 regstar_T rst;
4797
4798 /*
4799 * Lookahead to avoid useless match attempts when we know
4800 * what character comes next.
4801 */
4802 if (OP(next) == EXACTLY)
4803 {
4804 rst.nextb = *OPERAND(next);
4805 if (ireg_ic)
4806 {
4807 if (MB_ISUPPER(rst.nextb))
4808 rst.nextb_ic = MB_TOLOWER(rst.nextb);
4809 else
4810 rst.nextb_ic = MB_TOUPPER(rst.nextb);
4811 }
4812 else
4813 rst.nextb_ic = rst.nextb;
4814 }
4815 else
4816 {
4817 rst.nextb = NUL;
4818 rst.nextb_ic = NUL;
4819 }
4820 if (op != BRACE_SIMPLE)
4821 {
4822 rst.minval = (op == STAR) ? 0 : 1;
4823 rst.maxval = MAX_LIMIT;
4824 }
4825 else
4826 {
4827 rst.minval = bl_minval;
4828 rst.maxval = bl_maxval;
4829 }
4830
4831 /*
4832 * When maxval > minval, try matching as much as possible, up
4833 * to maxval. When maxval < minval, try matching at least the
4834 * minimal number (since the range is backwards, that's also
4835 * maxval!).
4836 */
4837 rst.count = regrepeat(OPERAND(scan), rst.maxval);
4838 if (got_int)
4839 {
4840 status = RA_FAIL;
4841 break;
4842 }
4843 if (rst.minval <= rst.maxval
4844 ? rst.count >= rst.minval : rst.count >= rst.maxval)
4845 {
4846 /* It could match. Prepare for trying to match what
4847 * follows. The code is below. Parameters are stored in
4848 * a regstar_T on the regstack. */
4849 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
4850 {
4851 EMSG(_(e_maxmempat));
4852 status = RA_FAIL;
4853 }
4854 else if (ga_grow(&regstack, sizeof(regstar_T)) == FAIL)
4855 status = RA_FAIL;
4856 else
4857 {
4858 regstack.ga_len += sizeof(regstar_T);
4859 rp = regstack_push(rst.minval <= rst.maxval
4860 ? RS_STAR_LONG : RS_STAR_SHORT, scan);
4861 if (rp == NULL)
4862 status = RA_FAIL;
4863 else
4864 {
4865 *(((regstar_T *)rp) - 1) = rst;
4866 status = RA_BREAK; /* skip the restore bits */
4867 }
4868 }
4869 }
4870 else
4871 status = RA_NOMATCH;
4872
4873 }
4874 break;
4875
4876 case NOMATCH:
4877 case MATCH:
4878 case SUBPAT:
4879 rp = regstack_push(RS_NOMATCH, scan);
4880 if (rp == NULL)
4881 status = RA_FAIL;
4882 else
4883 {
4884 rp->rs_no = op;
4885 reg_save(&rp->rs_un.regsave, &backpos);
4886 next = OPERAND(scan);
4887 /* We continue and handle the result when done. */
4888 }
4889 break;
4890
4891 case BEHIND:
4892 case NOBEHIND:
4893 /* Need a bit of room to store extra positions. */
4894 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
4895 {
4896 EMSG(_(e_maxmempat));
4897 status = RA_FAIL;
4898 }
4899 else if (ga_grow(&regstack, sizeof(regbehind_T)) == FAIL)
4900 status = RA_FAIL;
4901 else
4902 {
4903 regstack.ga_len += sizeof(regbehind_T);
4904 rp = regstack_push(RS_BEHIND1, scan);
4905 if (rp == NULL)
4906 status = RA_FAIL;
4907 else
4908 {
4909 /* Need to save the subexpr to be able to restore them
4910 * when there is a match but we don't use it. */
4911 save_subexpr(((regbehind_T *)rp) - 1);
4912
4913 rp->rs_no = op;
4914 reg_save(&rp->rs_un.regsave, &backpos);
4915 /* First try if what follows matches. If it does then we
4916 * check the behind match by looping. */
4917 }
4918 }
4919 break;
4920
4921 case BHPOS:
4922 if (REG_MULTI)
4923 {
4924 if (behind_pos.rs_u.pos.col != (colnr_T)(reginput - regline)
4925 || behind_pos.rs_u.pos.lnum != reglnum)
4926 status = RA_NOMATCH;
4927 }
4928 else if (behind_pos.rs_u.ptr != reginput)
4929 status = RA_NOMATCH;
4930 break;
4931
4932 case NEWL:
4933 if ((c != NUL || !REG_MULTI || reglnum > reg_maxline
4934 || reg_line_lbr) && (c != '\n' || !reg_line_lbr))
4935 status = RA_NOMATCH;
4936 else if (reg_line_lbr)
4937 ADVANCE_REGINPUT();
4938 else
4939 reg_nextline();
4940 break;
4941
4942 case END:
4943 status = RA_MATCH; /* Success! */
4944 break;
4945
4946 default:
4947 EMSG(_(e_re_corr));
4948 #ifdef DEBUG
4949 printf("Illegal op code %d\n", op);
4950 #endif
4951 status = RA_FAIL;
4952 break;
4953 }
4954 }
4955
4956 /* If we can't continue sequentially, break the inner loop. */
4957 if (status != RA_CONT)
4958 break;
4959
4960 /* Continue in inner loop, advance to next item. */
4961 scan = next;
4962
4963 } /* end of inner loop */
4964
4965 /*
4966 * If there is something on the regstack execute the code for the state.
4967 * If the state is popped then loop and use the older state.
4968 */
4969 while (regstack.ga_len > 0 && status != RA_FAIL)
4970 {
4971 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
4972 switch (rp->rs_state)
4973 {
4974 case RS_NOPEN:
4975 /* Result is passed on as-is, simply pop the state. */
4976 regstack_pop(&scan);
4977 break;
4978
4979 case RS_MOPEN:
4980 /* Pop the state. Restore pointers when there is no match. */
4981 if (status == RA_NOMATCH)
4982 restore_se(&rp->rs_un.sesave, &reg_startpos[rp->rs_no],
4983 &reg_startp[rp->rs_no]);
4984 regstack_pop(&scan);
4985 break;
4986
4987 #ifdef FEAT_SYN_HL
4988 case RS_ZOPEN:
4989 /* Pop the state. Restore pointers when there is no match. */
4990 if (status == RA_NOMATCH)
4991 restore_se(&rp->rs_un.sesave, &reg_startzpos[rp->rs_no],
4992 &reg_startzp[rp->rs_no]);
4993 regstack_pop(&scan);
4994 break;
4995 #endif
4996
4997 case RS_MCLOSE:
4998 /* Pop the state. Restore pointers when there is no match. */
4999 if (status == RA_NOMATCH)
5000 restore_se(&rp->rs_un.sesave, &reg_endpos[rp->rs_no],
5001 &reg_endp[rp->rs_no]);
5002 regstack_pop(&scan);
5003 break;
5004
5005 #ifdef FEAT_SYN_HL
5006 case RS_ZCLOSE:
5007 /* Pop the state. Restore pointers when there is no match. */
5008 if (status == RA_NOMATCH)
5009 restore_se(&rp->rs_un.sesave, &reg_endzpos[rp->rs_no],
5010 &reg_endzp[rp->rs_no]);
5011 regstack_pop(&scan);
5012 break;
5013 #endif
5014
5015 case RS_BRANCH:
5016 if (status == RA_MATCH)
5017 /* this branch matched, use it */
5018 regstack_pop(&scan);
5019 else
5020 {
5021 if (status != RA_BREAK)
5022 {
5023 /* After a non-matching branch: try next one. */
5024 reg_restore(&rp->rs_un.regsave, &backpos);
5025 scan = rp->rs_scan;
5026 }
5027 if (scan == NULL || OP(scan) != BRANCH)
5028 {
5029 /* no more branches, didn't find a match */
5030 status = RA_NOMATCH;
5031 regstack_pop(&scan);
5032 }
5033 else
5034 {
5035 /* Prepare to try a branch. */
5036 rp->rs_scan = regnext(scan);
5037 reg_save(&rp->rs_un.regsave, &backpos);
5038 scan = OPERAND(scan);
5039 }
5040 }
5041 break;
5042
5043 case RS_BRCPLX_MORE:
5044 /* Pop the state. Restore pointers when there is no match. */
5045 if (status == RA_NOMATCH)
5046 {
5047 reg_restore(&rp->rs_un.regsave, &backpos);
5048 --brace_count[rp->rs_no]; /* decrement match count */
5049 }
5050 regstack_pop(&scan);
5051 break;
5052
5053 case RS_BRCPLX_LONG:
5054 /* Pop the state. Restore pointers when there is no match. */
5055 if (status == RA_NOMATCH)
5056 {
5057 /* There was no match, but we did find enough matches. */
5058 reg_restore(&rp->rs_un.regsave, &backpos);
5059 --brace_count[rp->rs_no];
5060 /* continue with the items after "\{}" */
5061 status = RA_CONT;
5062 }
5063 regstack_pop(&scan);
5064 if (status == RA_CONT)
5065 scan = regnext(scan);
5066 break;
5067
5068 case RS_BRCPLX_SHORT:
5069 /* Pop the state. Restore pointers when there is no match. */
5070 if (status == RA_NOMATCH)
5071 /* There was no match, try to match one more item. */
5072 reg_restore(&rp->rs_un.regsave, &backpos);
5073 regstack_pop(&scan);
5074 if (status == RA_NOMATCH)
5075 {
5076 scan = OPERAND(scan);
5077 status = RA_CONT;
5078 }
5079 break;
5080
5081 case RS_NOMATCH:
5082 /* Pop the state. If the operand matches for NOMATCH or
5083 * doesn't match for MATCH/SUBPAT, we fail. Otherwise backup,
5084 * except for SUBPAT, and continue with the next item. */
5085 if (status == (rp->rs_no == NOMATCH ? RA_MATCH : RA_NOMATCH))
5086 status = RA_NOMATCH;
5087 else
5088 {
5089 status = RA_CONT;
5090 if (rp->rs_no != SUBPAT) /* zero-width */
5091 reg_restore(&rp->rs_un.regsave, &backpos);
5092 }
5093 regstack_pop(&scan);
5094 if (status == RA_CONT)
5095 scan = regnext(scan);
5096 break;
5097
5098 case RS_BEHIND1:
5099 if (status == RA_NOMATCH)
5100 {
5101 regstack_pop(&scan);
5102 regstack.ga_len -= sizeof(regbehind_T);
5103 }
5104 else
5105 {
5106 /* The stuff after BEHIND/NOBEHIND matches. Now try if
5107 * the behind part does (not) match before the current
5108 * position in the input. This must be done at every
5109 * position in the input and checking if the match ends at
5110 * the current position. */
5111
5112 /* save the position after the found match for next */
5113 reg_save(&(((regbehind_T *)rp) - 1)->save_after, &backpos);
5114
5115 /* start looking for a match with operand at the current
5116 * position. Go back one character until we find the
5117 * result, hitting the start of the line or the previous
5118 * line (for multi-line matching).
5119 * Set behind_pos to where the match should end, BHPOS
5120 * will match it. Save the current value. */
5121 (((regbehind_T *)rp) - 1)->save_behind = behind_pos;
5122 behind_pos = rp->rs_un.regsave;
5123
5124 rp->rs_state = RS_BEHIND2;
5125
5126 reg_restore(&rp->rs_un.regsave, &backpos);
5127 scan = OPERAND(rp->rs_scan);
5128 }
5129 break;
5130
5131 case RS_BEHIND2:
5132 /*
5133 * Looping for BEHIND / NOBEHIND match.
5134 */
5135 if (status == RA_MATCH && reg_save_equal(&behind_pos))
5136 {
5137 /* found a match that ends where "next" started */
5138 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
5139 if (rp->rs_no == BEHIND)
5140 reg_restore(&(((regbehind_T *)rp) - 1)->save_after,
5141 &backpos);
5142 else
5143 {
5144 /* But we didn't want a match. Need to restore the
5145 * subexpr, because what follows matched, so they have
5146 * been set. */
5147 status = RA_NOMATCH;
5148 restore_subexpr(((regbehind_T *)rp) - 1);
5149 }
5150 regstack_pop(&scan);
5151 regstack.ga_len -= sizeof(regbehind_T);
5152 }
5153 else
5154 {
5155 /* No match or a match that doesn't end where we want it: Go
5156 * back one character. May go to previous line once. */
5157 no = OK;
5158 if (REG_MULTI)
5159 {
5160 if (rp->rs_un.regsave.rs_u.pos.col == 0)
5161 {
5162 if (rp->rs_un.regsave.rs_u.pos.lnum
5163 < behind_pos.rs_u.pos.lnum
5164 || reg_getline(
5165 --rp->rs_un.regsave.rs_u.pos.lnum)
5166 == NULL)
5167 no = FAIL;
5168 else
5169 {
5170 reg_restore(&rp->rs_un.regsave, &backpos);
5171 rp->rs_un.regsave.rs_u.pos.col =
5172 (colnr_T)STRLEN(regline);
5173 }
5174 }
5175 else
5176 --rp->rs_un.regsave.rs_u.pos.col;
5177 }
5178 else
5179 {
5180 if (rp->rs_un.regsave.rs_u.ptr == regline)
5181 no = FAIL;
5182 else
5183 --rp->rs_un.regsave.rs_u.ptr;
5184 }
5185 if (no == OK)
5186 {
5187 /* Advanced, prepare for finding match again. */
5188 reg_restore(&rp->rs_un.regsave, &backpos);
5189 scan = OPERAND(rp->rs_scan);
5190 if (status == RA_MATCH)
5191 {
5192 /* We did match, so subexpr may have been changed,
5193 * need to restore them for the next try. */
5194 status = RA_NOMATCH;
5195 restore_subexpr(((regbehind_T *)rp) - 1);
5196 }
5197 }
5198 else
5199 {
5200 /* Can't advance. For NOBEHIND that's a match. */
5201 behind_pos = (((regbehind_T *)rp) - 1)->save_behind;
5202 if (rp->rs_no == NOBEHIND)
5203 {
5204 reg_restore(&(((regbehind_T *)rp) - 1)->save_after,
5205 &backpos);
5206 status = RA_MATCH;
5207 }
5208 else
5209 {
5210 /* We do want a proper match. Need to restore the
5211 * subexpr if we had a match, because they may have
5212 * been set. */
5213 if (status == RA_MATCH)
5214 {
5215 status = RA_NOMATCH;
5216 restore_subexpr(((regbehind_T *)rp) - 1);
5217 }
5218 }
5219 regstack_pop(&scan);
5220 regstack.ga_len -= sizeof(regbehind_T);
5221 }
5222 }
5223 break;
5224
5225 case RS_STAR_LONG:
5226 case RS_STAR_SHORT:
5227 {
5228 regstar_T *rst = ((regstar_T *)rp) - 1;
5229
5230 if (status == RA_MATCH)
5231 {
5232 regstack_pop(&scan);
5233 regstack.ga_len -= sizeof(regstar_T);
5234 break;
5235 }
5236
5237 /* Tried once already, restore input pointers. */
5238 if (status != RA_BREAK)
5239 reg_restore(&rp->rs_un.regsave, &backpos);
5240
5241 /* Repeat until we found a position where it could match. */
5242 for (;;)
5243 {
5244 if (status != RA_BREAK)
5245 {
5246 /* Tried first position already, advance. */
5247 if (rp->rs_state == RS_STAR_LONG)
5248 {
5249 /* Trying for longest match, but couldn't or
5250 * didn't match -- back up one char. */
5251 if (--rst->count < rst->minval)
5252 break;
5253 if (reginput == regline)
5254 {
5255 /* backup to last char of previous line */
5256 --reglnum;
5257 regline = reg_getline(reglnum);
5258 /* Just in case regrepeat() didn't count
5259 * right. */
5260 if (regline == NULL)
5261 break;
5262 reginput = regline + STRLEN(regline);
5263 fast_breakcheck();
5264 }
5265 else
5266 mb_ptr_back(regline, reginput);
5267 }
5268 else
5269 {
5270 /* Range is backwards, use shortest match first.
5271 * Careful: maxval and minval are exchanged!
5272 * Couldn't or didn't match: try advancing one
5273 * char. */
5274 if (rst->count == rst->minval
5275 || regrepeat(OPERAND(rp->rs_scan), 1L) == 0)
5276 break;
5277 ++rst->count;
5278 }
5279 if (got_int)
5280 break;
5281 }
5282 else
5283 status = RA_NOMATCH;
5284
5285 /* If it could match, try it. */
5286 if (rst->nextb == NUL || *reginput == rst->nextb
5287 || *reginput == rst->nextb_ic)
5288 {
5289 reg_save(&rp->rs_un.regsave, &backpos);
5290 scan = regnext(rp->rs_scan);
5291 status = RA_CONT;
5292 break;
5293 }
5294 }
5295 if (status != RA_CONT)
5296 {
5297 /* Failed. */
5298 regstack_pop(&scan);
5299 regstack.ga_len -= sizeof(regstar_T);
5300 status = RA_NOMATCH;
5301 }
5302 }
5303 break;
5304 }
5305
5306 /* If we want to continue the inner loop or didn't pop a state
5307 * continue matching loop */
5308 if (status == RA_CONT || rp == (regitem_T *)
5309 ((char *)regstack.ga_data + regstack.ga_len) - 1)
5310 break;
5311 }
5312
5313 /* May need to continue with the inner loop, starting at "scan". */
5314 if (status == RA_CONT)
5315 continue;
5316
5317 /*
5318 * If the regstack is empty or something failed we are done.
5319 */
5320 if (regstack.ga_len == 0 || status == RA_FAIL)
5321 {
5322 if (scan == NULL)
5323 {
5324 /*
5325 * We get here only if there's trouble -- normally "case END" is
5326 * the terminating point.
5327 */
5328 EMSG(_(e_re_corr));
5329 #ifdef DEBUG
5330 printf("Premature EOL\n");
5331 #endif
5332 }
5333 if (status == RA_FAIL)
5334 got_int = TRUE;
5335 return (status == RA_MATCH);
5336 }
5337
5338 } /* End of loop until the regstack is empty. */
5339
5340 /* NOTREACHED */
5341 }
5342
5343 /*
5344 * Push an item onto the regstack.
5345 * Returns pointer to new item. Returns NULL when out of memory.
5346 */
5347 static regitem_T *
5348 regstack_push(state, scan)
5349 regstate_T state;
5350 char_u *scan;
5351 {
5352 regitem_T *rp;
5353
5354 if ((long)((unsigned)regstack.ga_len >> 10) >= p_mmp)
5355 {
5356 EMSG(_(e_maxmempat));
5357 return NULL;
5358 }
5359 if (ga_grow(&regstack, sizeof(regitem_T)) == FAIL)
5360 return NULL;
5361
5362 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len);
5363 rp->rs_state = state;
5364 rp->rs_scan = scan;
5365
5366 regstack.ga_len += sizeof(regitem_T);
5367 return rp;
5368 }
5369
5370 /*
5371 * Pop an item from the regstack.
5372 */
5373 static void
5374 regstack_pop(scan)
5375 char_u **scan;
5376 {
5377 regitem_T *rp;
5378
5379 rp = (regitem_T *)((char *)regstack.ga_data + regstack.ga_len) - 1;
5380 *scan = rp->rs_scan;
5381
5382 regstack.ga_len -= sizeof(regitem_T);
5383 }
5384
5385 /*
5386 * regrepeat - repeatedly match something simple, return how many.
5387 * Advances reginput (and reglnum) to just after the matched chars.
5388 */
5389 static int
5390 regrepeat(p, maxcount)
5391 char_u *p;
5392 long maxcount; /* maximum number of matches allowed */
5393 {
5394 long count = 0;
5395 char_u *scan;
5396 char_u *opnd;
5397 int mask;
5398 int testval = 0;
5399
5400 scan = reginput; /* Make local copy of reginput for speed. */
5401 opnd = OPERAND(p);
5402 switch (OP(p))
5403 {
5404 case ANY:
5405 case ANY + ADD_NL:
5406 while (count < maxcount)
5407 {
5408 /* Matching anything means we continue until end-of-line (or
5409 * end-of-file for ANY + ADD_NL), only limited by maxcount. */
5410 while (*scan != NUL && count < maxcount)
5411 {
5412 ++count;
5413 mb_ptr_adv(scan);
5414 }
5415 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline
5416 || reg_line_lbr || count == maxcount)
5417 break;
5418 ++count; /* count the line-break */
5419 reg_nextline();
5420 scan = reginput;
5421 if (got_int)
5422 break;
5423 }
5424 break;
5425
5426 case IDENT:
5427 case IDENT + ADD_NL:
5428 testval = TRUE;
5429 /*FALLTHROUGH*/
5430 case SIDENT:
5431 case SIDENT + ADD_NL:
5432 while (count < maxcount)
5433 {
5434 if (vim_isIDc(*scan) && (testval || !VIM_ISDIGIT(*scan)))
5435 {
5436 mb_ptr_adv(scan);
5437 }
5438 else if (*scan == NUL)
5439 {
5440 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline
5441 || reg_line_lbr)
5442 break;
5443 reg_nextline();
5444 scan = reginput;
5445 if (got_int)
5446 break;
5447 }
5448 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
5449 ++scan;
5450 else
5451 break;
5452 ++count;
5453 }
5454 break;
5455
5456 case KWORD:
5457 case KWORD + ADD_NL:
5458 testval = TRUE;
5459 /*FALLTHROUGH*/
5460 case SKWORD:
5461 case SKWORD + ADD_NL:
5462 while (count < maxcount)
5463 {
5464 if (vim_iswordp(scan) && (testval || !VIM_ISDIGIT(*scan)))
5465 {
5466 mb_ptr_adv(scan);
5467 }
5468 else if (*scan == NUL)
5469 {
5470 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline
5471 || reg_line_lbr)
5472 break;
5473 reg_nextline();
5474 scan = reginput;
5475 if (got_int)
5476 break;
5477 }
5478 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
5479 ++scan;
5480 else
5481 break;
5482 ++count;
5483 }
5484 break;
5485
5486 case FNAME:
5487 case FNAME + ADD_NL:
5488 testval = TRUE;
5489 /*FALLTHROUGH*/
5490 case SFNAME:
5491 case SFNAME + ADD_NL:
5492 while (count < maxcount)
5493 {
5494 if (vim_isfilec(*scan) && (testval || !VIM_ISDIGIT(*scan)))
5495 {
5496 mb_ptr_adv(scan);
5497 }
5498 else if (*scan == NUL)
5499 {
5500 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline
5501 || reg_line_lbr)
5502 break;
5503 reg_nextline();
5504 scan = reginput;
5505 if (got_int)
5506 break;
5507 }
5508 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
5509 ++scan;
5510 else
5511 break;
5512 ++count;
5513 }
5514 break;
5515
5516 case PRINT:
5517 case PRINT + ADD_NL:
5518 testval = TRUE;
5519 /*FALLTHROUGH*/
5520 case SPRINT:
5521 case SPRINT + ADD_NL:
5522 while (count < maxcount)
5523 {
5524 if (*scan == NUL)
5525 {
5526 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline
5527 || reg_line_lbr)
5528 break;
5529 reg_nextline();
5530 scan = reginput;
5531 if (got_int)
5532 break;
5533 }
5534 else if (ptr2cells(scan) == 1 && (testval || !VIM_ISDIGIT(*scan)))
5535 {
5536 mb_ptr_adv(scan);
5537 }
5538 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
5539 ++scan;
5540 else
5541 break;
5542 ++count;
5543 }
5544 break;
5545
5546 case WHITE:
5547 case WHITE + ADD_NL:
5548 testval = mask = RI_WHITE;
5549 do_class:
5550 while (count < maxcount)
5551 {
5552 #ifdef FEAT_MBYTE
5553 int l;
5554 #endif
5555 if (*scan == NUL)
5556 {
5557 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline
5558 || reg_line_lbr)
5559 break;
5560 reg_nextline();
5561 scan = reginput;
5562 if (got_int)
5563 break;
5564 }
5565 #ifdef FEAT_MBYTE
5566 else if (has_mbyte && (l = (*mb_ptr2len)(scan)) > 1)
5567 {
5568 if (testval != 0)
5569 break;
5570 scan += l;
5571 }
5572 #endif
5573 else if ((class_tab[*scan] & mask) == testval)
5574 ++scan;
5575 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
5576 ++scan;
5577 else
5578 break;
5579 ++count;
5580 }
5581 break;
5582
5583 case NWHITE:
5584 case NWHITE + ADD_NL:
5585 mask = RI_WHITE;
5586 goto do_class;
5587 case DIGIT:
5588 case DIGIT + ADD_NL:
5589 testval = mask = RI_DIGIT;
5590 goto do_class;
5591 case NDIGIT:
5592 case NDIGIT + ADD_NL:
5593 mask = RI_DIGIT;
5594 goto do_class;
5595 case HEX:
5596 case HEX + ADD_NL:
5597 testval = mask = RI_HEX;
5598 goto do_class;
5599 case NHEX:
5600 case NHEX + ADD_NL:
5601 mask = RI_HEX;
5602 goto do_class;
5603 case OCTAL:
5604 case OCTAL + ADD_NL:
5605 testval = mask = RI_OCTAL;
5606 goto do_class;
5607 case NOCTAL:
5608 case NOCTAL + ADD_NL:
5609 mask = RI_OCTAL;
5610 goto do_class;
5611 case WORD:
5612 case WORD + ADD_NL:
5613 testval = mask = RI_WORD;
5614 goto do_class;
5615 case NWORD:
5616 case NWORD + ADD_NL:
5617 mask = RI_WORD;
5618 goto do_class;
5619 case HEAD:
5620 case HEAD + ADD_NL:
5621 testval = mask = RI_HEAD;
5622 goto do_class;
5623 case NHEAD:
5624 case NHEAD + ADD_NL:
5625 mask = RI_HEAD;
5626 goto do_class;
5627 case ALPHA:
5628 case ALPHA + ADD_NL:
5629 testval = mask = RI_ALPHA;
5630 goto do_class;
5631 case NALPHA:
5632 case NALPHA + ADD_NL:
5633 mask = RI_ALPHA;
5634 goto do_class;
5635 case LOWER:
5636 case LOWER + ADD_NL:
5637 testval = mask = RI_LOWER;
5638 goto do_class;
5639 case NLOWER:
5640 case NLOWER + ADD_NL:
5641 mask = RI_LOWER;
5642 goto do_class;
5643 case UPPER:
5644 case UPPER + ADD_NL:
5645 testval = mask = RI_UPPER;
5646 goto do_class;
5647 case NUPPER:
5648 case NUPPER + ADD_NL:
5649 mask = RI_UPPER;
5650 goto do_class;
5651
5652 case EXACTLY:
5653 {
5654 int cu, cl;
5655
5656 /* This doesn't do a multi-byte character, because a MULTIBYTECODE
5657 * would have been used for it. It does handle single-byte
5658 * characters, such as latin1. */
5659 if (ireg_ic)
5660 {
5661 cu = MB_TOUPPER(*opnd);
5662 cl = MB_TOLOWER(*opnd);
5663 while (count < maxcount && (*scan == cu || *scan == cl))
5664 {
5665 count++;
5666 scan++;
5667 }
5668 }
5669 else
5670 {
5671 cu = *opnd;
5672 while (count < maxcount && *scan == cu)
5673 {
5674 count++;
5675 scan++;
5676 }
5677 }
5678 break;
5679 }
5680
5681 #ifdef FEAT_MBYTE
5682 case MULTIBYTECODE:
5683 {
5684 int i, len, cf = 0;
5685
5686 /* Safety check (just in case 'encoding' was changed since
5687 * compiling the program). */
5688 if ((len = (*mb_ptr2len)(opnd)) > 1)
5689 {
5690 if (ireg_ic && enc_utf8)
5691 cf = utf_fold(utf_ptr2char(opnd));
5692 while (count < maxcount)
5693 {
5694 for (i = 0; i < len; ++i)
5695 if (opnd[i] != scan[i])
5696 break;
5697 if (i < len && (!ireg_ic || !enc_utf8
5698 || utf_fold(utf_ptr2char(scan)) != cf))
5699 break;
5700 scan += len;
5701 ++count;
5702 }
5703 }
5704 }
5705 break;
5706 #endif
5707
5708 case ANYOF:
5709 case ANYOF + ADD_NL:
5710 testval = TRUE;
5711 /*FALLTHROUGH*/
5712
5713 case ANYBUT:
5714 case ANYBUT + ADD_NL:
5715 while (count < maxcount)
5716 {
5717 #ifdef FEAT_MBYTE
5718 int len;
5719 #endif
5720 if (*scan == NUL)
5721 {
5722 if (!REG_MULTI || !WITH_NL(OP(p)) || reglnum > reg_maxline
5723 || reg_line_lbr)
5724 break;
5725 reg_nextline();
5726 scan = reginput;
5727 if (got_int)
5728 break;
5729 }
5730 else if (reg_line_lbr && *scan == '\n' && WITH_NL(OP(p)))
5731 ++scan;
5732 #ifdef FEAT_MBYTE
5733 else if (has_mbyte && (len = (*mb_ptr2len)(scan)) > 1)
5734 {
5735 if ((cstrchr(opnd, (*mb_ptr2char)(scan)) == NULL) == testval)
5736 break;
5737 scan += len;
5738 }
5739 #endif
5740 else
5741 {
5742 if ((cstrchr(opnd, *scan) == NULL) == testval)
5743 break;
5744 ++scan;
5745 }
5746 ++count;
5747 }
5748 break;
5749
5750 case NEWL:
5751 while (count < maxcount
5752 && ((*scan == NUL && reglnum <= reg_maxline && !reg_line_lbr
5753 && REG_MULTI) || (*scan == '\n' && reg_line_lbr)))
5754 {
5755 count++;
5756 if (reg_line_lbr)
5757 ADVANCE_REGINPUT();
5758 else
5759 reg_nextline();
5760 scan = reginput;
5761 if (got_int)
5762 break;
5763 }
5764 break;
5765
5766 default: /* Oh dear. Called inappropriately. */
5767 EMSG(_(e_re_corr));
5768 #ifdef DEBUG
5769 printf("Called regrepeat with op code %d\n", OP(p));
5770 #endif
5771 break;
5772 }
5773
5774 reginput = scan;
5775
5776 return (int)count;
5777 }
5778
5779 /*
5780 * regnext - dig the "next" pointer out of a node
5781 * Returns NULL when calculating size, when there is no next item and when
5782 * there is an error.
5783 */
5784 static char_u *
5785 regnext(p)
5786 char_u *p;
5787 {
5788 int offset;
5789
5790 if (p == JUST_CALC_SIZE || reg_toolong)
5791 return NULL;
5792
5793 offset = NEXT(p);
5794 if (offset == 0)
5795 return NULL;
5796
5797 if (OP(p) == BACK)
5798 return p - offset;
5799 else
5800 return p + offset;
5801 }
5802
5803 /*
5804 * Check the regexp program for its magic number.
5805 * Return TRUE if it's wrong.
5806 */
5807 static int
5808 prog_magic_wrong()
5809 {
5810 if (UCHARAT(REG_MULTI
5811 ? reg_mmatch->regprog->program
5812 : reg_match->regprog->program) != REGMAGIC)
5813 {
5814 EMSG(_(e_re_corr));
5815 return TRUE;
5816 }
5817 return FALSE;
5818 }
5819
5820 /*
5821 * Cleanup the subexpressions, if this wasn't done yet.
5822 * This construction is used to clear the subexpressions only when they are
5823 * used (to increase speed).
5824 */
5825 static void
5826 cleanup_subexpr()
5827 {
5828 if (need_clear_subexpr)
5829 {
5830 if (REG_MULTI)
5831 {
5832 /* Use 0xff to set lnum to -1 */
5833 vim_memset(reg_startpos, 0xff, sizeof(lpos_T) * NSUBEXP);
5834 vim_memset(reg_endpos, 0xff, sizeof(lpos_T) * NSUBEXP);
5835 }
5836 else
5837 {
5838 vim_memset(reg_startp, 0, sizeof(char_u *) * NSUBEXP);
5839 vim_memset(reg_endp, 0, sizeof(char_u *) * NSUBEXP);
5840 }
5841 need_clear_subexpr = FALSE;
5842 }
5843 }
5844
5845 #ifdef FEAT_SYN_HL
5846 static void
5847 cleanup_zsubexpr()
5848 {
5849 if (need_clear_zsubexpr)
5850 {
5851 if (REG_MULTI)
5852 {
5853 /* Use 0xff to set lnum to -1 */
5854 vim_memset(reg_startzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
5855 vim_memset(reg_endzpos, 0xff, sizeof(lpos_T) * NSUBEXP);
5856 }
5857 else
5858 {
5859 vim_memset(reg_startzp, 0, sizeof(char_u *) * NSUBEXP);
5860 vim_memset(reg_endzp, 0, sizeof(char_u *) * NSUBEXP);
5861 }
5862 need_clear_zsubexpr = FALSE;
5863 }
5864 }
5865 #endif
5866
5867 /*
5868 * Save the current subexpr to "bp", so that they can be restored
5869 * later by restore_subexpr().
5870 */
5871 static void
5872 save_subexpr(bp)
5873 regbehind_T *bp;
5874 {
5875 int i;
5876
5877 /* When "need_clear_subexpr" is set we don't need to save the values, only
5878 * remember that this flag needs to be set again when restoring. */
5879 bp->save_need_clear_subexpr = need_clear_subexpr;
5880 if (!need_clear_subexpr)
5881 {
5882 for (i = 0; i < NSUBEXP; ++i)
5883 {
5884 if (REG_MULTI)
5885 {
5886 bp->save_start[i].se_u.pos = reg_startpos[i];
5887 bp->save_end[i].se_u.pos = reg_endpos[i];
5888 }
5889 else
5890 {
5891 bp->save_start[i].se_u.ptr = reg_startp[i];
5892 bp->save_end[i].se_u.ptr = reg_endp[i];
5893 }
5894 }
5895 }
5896 }
5897
5898 /*
5899 * Restore the subexpr from "bp".
5900 */
5901 static void
5902 restore_subexpr(bp)
5903 regbehind_T *bp;
5904 {
5905 int i;
5906
5907 /* Only need to restore saved values when they are not to be cleared. */
5908 need_clear_subexpr = bp->save_need_clear_subexpr;
5909 if (!need_clear_subexpr)
5910 {
5911 for (i = 0; i < NSUBEXP; ++i)
5912 {
5913 if (REG_MULTI)
5914 {
5915 reg_startpos[i] = bp->save_start[i].se_u.pos;
5916 reg_endpos[i] = bp->save_end[i].se_u.pos;
5917 }
5918 else
5919 {
5920 reg_startp[i] = bp->save_start[i].se_u.ptr;
5921 reg_endp[i] = bp->save_end[i].se_u.ptr;
5922 }
5923 }
5924 }
5925 }
5926
5927 /*
5928 * Advance reglnum, regline and reginput to the next line.
5929 */
5930 static void
5931 reg_nextline()
5932 {
5933 regline = reg_getline(++reglnum);
5934 reginput = regline;
5935 fast_breakcheck();
5936 }
5937
5938 /*
5939 * Save the input line and position in a regsave_T.
5940 */
5941 static void
5942 reg_save(save, gap)
5943 regsave_T *save;
5944 garray_T *gap;
5945 {
5946 if (REG_MULTI)
5947 {
5948 save->rs_u.pos.col = (colnr_T)(reginput - regline);
5949 save->rs_u.pos.lnum = reglnum;
5950 }
5951 else
5952 save->rs_u.ptr = reginput;
5953 save->rs_len = gap->ga_len;
5954 }
5955
5956 /*
5957 * Restore the input line and position from a regsave_T.
5958 */
5959 static void
5960 reg_restore(save, gap)
5961 regsave_T *save;
5962 garray_T *gap;
5963 {
5964 if (REG_MULTI)
5965 {
5966 if (reglnum != save->rs_u.pos.lnum)
5967 {
5968 /* only call reg_getline() when the line number changed to save
5969 * a bit of time */
5970 reglnum = save->rs_u.pos.lnum;
5971 regline = reg_getline(reglnum);
5972 }
5973 reginput = regline + save->rs_u.pos.col;
5974 }
5975 else
5976 reginput = save->rs_u.ptr;
5977 gap->ga_len = save->rs_len;
5978 }
5979
5980 /*
5981 * Return TRUE if current position is equal to saved position.
5982 */
5983 static int
5984 reg_save_equal(save)
5985 regsave_T *save;
5986 {
5987 if (REG_MULTI)
5988 return reglnum == save->rs_u.pos.lnum
5989 && reginput == regline + save->rs_u.pos.col;
5990 return reginput == save->rs_u.ptr;
5991 }
5992
5993 /*
5994 * Tentatively set the sub-expression start to the current position (after
5995 * calling regmatch() they will have changed). Need to save the existing
5996 * values for when there is no match.
5997 * Use se_save() to use pointer (save_se_multi()) or position (save_se_one()),
5998 * depending on REG_MULTI.
5999 */
6000 static void
6001 save_se_multi(savep, posp)
6002 save_se_T *savep;
6003 lpos_T *posp;
6004 {
6005 savep->se_u.pos = *posp;
6006 posp->lnum = reglnum;
6007 posp->col = (colnr_T)(reginput - regline);
6008 }
6009
6010 static void
6011 save_se_one(savep, pp)
6012 save_se_T *savep;
6013 char_u **pp;
6014 {
6015 savep->se_u.ptr = *pp;
6016 *pp = reginput;
6017 }
6018
6019 /*
6020 * Compare a number with the operand of RE_LNUM, RE_COL or RE_VCOL.
6021 */
6022 static int
6023 re_num_cmp(val, scan)
6024 long_u val;
6025 char_u *scan;
6026 {
6027 long_u n = OPERAND_MIN(scan);
6028
6029 if (OPERAND_CMP(scan) == '>')
6030 return val > n;
6031 if (OPERAND_CMP(scan) == '<')
6032 return val < n;
6033 return val == n;
6034 }
6035
6036
6037 #ifdef DEBUG
6038
6039 /*
6040 * regdump - dump a regexp onto stdout in vaguely comprehensible form
6041 */
6042 static void
6043 regdump(pattern, r)
6044 char_u *pattern;
6045 regprog_T *r;
6046 {
6047 char_u *s;
6048 int op = EXACTLY; /* Arbitrary non-END op. */
6049 char_u *next;
6050 char_u *end = NULL;
6051
6052 printf("\r\nregcomp(%s):\r\n", pattern);
6053
6054 s = r->program + 1;
6055 /*
6056 * Loop until we find the END that isn't before a referred next (an END
6057 * can also appear in a NOMATCH operand).
6058 */
6059 while (op != END || s <= end)
6060 {
6061 op = OP(s);
6062 printf("%2d%s", (int)(s - r->program), regprop(s)); /* Where, what. */
6063 next = regnext(s);
6064 if (next == NULL) /* Next ptr. */
6065 printf("(0)");
6066 else
6067 printf("(%d)", (int)((s - r->program) + (next - s)));
6068 if (end < next)
6069 end = next;
6070 if (op == BRACE_LIMITS)
6071 {
6072 /* Two short ints */
6073 printf(" minval %ld, maxval %ld", OPERAND_MIN(s), OPERAND_MAX(s));
6074 s += 8;
6075 }
6076 s += 3;
6077 if (op == ANYOF || op == ANYOF + ADD_NL
6078 || op == ANYBUT || op == ANYBUT + ADD_NL
6079 || op == EXACTLY)
6080 {
6081 /* Literal string, where present. */
6082 while (*s != NUL)
6083 printf("%c", *s++);
6084 s++;
6085 }
6086 printf("\r\n");
6087 }
6088
6089 /* Header fields of interest. */
6090 if (r->regstart != NUL)
6091 printf("start `%s' 0x%x; ", r->regstart < 256
6092 ? (char *)transchar(r->regstart)
6093 : "multibyte", r->regstart);
6094 if (r->reganch)
6095 printf("anchored; ");
6096 if (r->regmust != NULL)
6097 printf("must have \"%s\"", r->regmust);
6098 printf("\r\n");
6099 }
6100
6101 /*
6102 * regprop - printable representation of opcode
6103 */
6104 static char_u *
6105 regprop(op)
6106 char_u *op;
6107 {
6108 char_u *p;
6109 static char_u buf[50];
6110
6111 (void) strcpy(buf, ":");
6112
6113 switch (OP(op))
6114 {
6115 case BOL:
6116 p = "BOL";
6117 break;
6118 case EOL:
6119 p = "EOL";
6120 break;
6121 case RE_BOF:
6122 p = "BOF";
6123 break;
6124 case RE_EOF:
6125 p = "EOF";
6126 break;
6127 case CURSOR:
6128 p = "CURSOR";
6129 break;
6130 case RE_VISUAL:
6131 p = "RE_VISUAL";
6132 break;
6133 case RE_LNUM:
6134 p = "RE_LNUM";
6135 break;
6136 case RE_MARK:
6137 p = "RE_MARK";
6138 break;
6139 case RE_COL:
6140 p = "RE_COL";
6141 break;
6142 case RE_VCOL:
6143 p = "RE_VCOL";
6144 break;
6145 case BOW:
6146 p = "BOW";
6147 break;
6148 case EOW:
6149 p = "EOW";
6150 break;
6151 case ANY:
6152 p = "ANY";
6153 break;
6154 case ANY + ADD_NL:
6155 p = "ANY+NL";
6156 break;
6157 case ANYOF:
6158 p = "ANYOF";
6159 break;
6160 case ANYOF + ADD_NL:
6161 p = "ANYOF+NL";
6162 break;
6163 case ANYBUT:
6164 p = "ANYBUT";
6165 break;
6166 case ANYBUT + ADD_NL:
6167 p = "ANYBUT+NL";
6168 break;
6169 case IDENT:
6170 p = "IDENT";
6171 break;
6172 case IDENT + ADD_NL:
6173 p = "IDENT+NL";
6174 break;
6175 case SIDENT:
6176 p = "SIDENT";
6177 break;
6178 case SIDENT + ADD_NL:
6179 p = "SIDENT+NL";
6180 break;
6181 case KWORD:
6182 p = "KWORD";
6183 break;
6184 case KWORD + ADD_NL:
6185 p = "KWORD+NL";
6186 break;
6187 case SKWORD:
6188 p = "SKWORD";
6189 break;
6190 case SKWORD + ADD_NL:
6191 p = "SKWORD+NL";
6192 break;
6193 case FNAME:
6194 p = "FNAME";
6195 break;
6196 case FNAME + ADD_NL:
6197 p = "FNAME+NL";
6198 break;
6199 case SFNAME:
6200 p = "SFNAME";
6201 break;
6202 case SFNAME + ADD_NL:
6203 p = "SFNAME+NL";
6204 break;
6205 case PRINT:
6206 p = "PRINT";
6207 break;
6208 case PRINT + ADD_NL:
6209 p = "PRINT+NL";
6210 break;
6211 case SPRINT:
6212 p = "SPRINT";
6213 break;
6214 case SPRINT + ADD_NL:
6215 p = "SPRINT+NL";
6216 break;
6217 case WHITE:
6218 p = "WHITE";
6219 break;
6220 case WHITE + ADD_NL:
6221 p = "WHITE+NL";
6222 break;
6223 case NWHITE:
6224 p = "NWHITE";
6225 break;
6226 case NWHITE + ADD_NL:
6227 p = "NWHITE+NL";
6228 break;
6229 case DIGIT:
6230 p = "DIGIT";
6231 break;
6232 case DIGIT + ADD_NL:
6233 p = "DIGIT+NL";
6234 break;
6235 case NDIGIT:
6236 p = "NDIGIT";
6237 break;
6238 case NDIGIT + ADD_NL:
6239 p = "NDIGIT+NL";
6240 break;
6241 case HEX:
6242 p = "HEX";
6243 break;
6244 case HEX + ADD_NL:
6245 p = "HEX+NL";
6246 break;
6247 case NHEX:
6248 p = "NHEX";
6249 break;
6250 case NHEX + ADD_NL:
6251 p = "NHEX+NL";
6252 break;
6253 case OCTAL:
6254 p = "OCTAL";
6255 break;
6256 case OCTAL + ADD_NL:
6257 p = "OCTAL+NL";
6258 break;
6259 case NOCTAL:
6260 p = "NOCTAL";
6261 break;
6262 case NOCTAL + ADD_NL:
6263 p = "NOCTAL+NL";
6264 break;
6265 case WORD:
6266 p = "WORD";
6267 break;
6268 case WORD + ADD_NL:
6269 p = "WORD+NL";
6270 break;
6271 case NWORD:
6272 p = "NWORD";
6273 break;
6274 case NWORD + ADD_NL:
6275 p = "NWORD+NL";
6276 break;
6277 case HEAD:
6278 p = "HEAD";
6279 break;
6280 case HEAD + ADD_NL:
6281 p = "HEAD+NL";
6282 break;
6283 case NHEAD:
6284 p = "NHEAD";
6285 break;
6286 case NHEAD + ADD_NL:
6287 p = "NHEAD+NL";
6288 break;
6289 case ALPHA:
6290 p = "ALPHA";
6291 break;
6292 case ALPHA + ADD_NL:
6293 p = "ALPHA+NL";
6294 break;
6295 case NALPHA:
6296 p = "NALPHA";
6297 break;
6298 case NALPHA + ADD_NL:
6299 p = "NALPHA+NL";
6300 break;
6301 case LOWER:
6302 p = "LOWER";
6303 break;
6304 case LOWER + ADD_NL:
6305 p = "LOWER+NL";
6306 break;
6307 case NLOWER:
6308 p = "NLOWER";
6309 break;
6310 case NLOWER + ADD_NL:
6311 p = "NLOWER+NL";
6312 break;
6313 case UPPER:
6314 p = "UPPER";
6315 break;
6316 case UPPER + ADD_NL:
6317 p = "UPPER+NL";
6318 break;
6319 case NUPPER:
6320 p = "NUPPER";
6321 break;
6322 case NUPPER + ADD_NL:
6323 p = "NUPPER+NL";
6324 break;
6325 case BRANCH:
6326 p = "BRANCH";
6327 break;
6328 case EXACTLY:
6329 p = "EXACTLY";
6330 break;
6331 case NOTHING:
6332 p = "NOTHING";
6333 break;
6334 case BACK:
6335 p = "BACK";
6336 break;
6337 case END:
6338 p = "END";
6339 break;
6340 case MOPEN + 0:
6341 p = "MATCH START";
6342 break;
6343 case MOPEN + 1:
6344 case MOPEN + 2:
6345 case MOPEN + 3:
6346 case MOPEN + 4:
6347 case MOPEN + 5:
6348 case MOPEN + 6:
6349 case MOPEN + 7:
6350 case MOPEN + 8:
6351 case MOPEN + 9:
6352 sprintf(buf + STRLEN(buf), "MOPEN%d", OP(op) - MOPEN);
6353 p = NULL;
6354 break;
6355 case MCLOSE + 0:
6356 p = "MATCH END";
6357 break;
6358 case MCLOSE + 1:
6359 case MCLOSE + 2:
6360 case MCLOSE + 3:
6361 case MCLOSE + 4:
6362 case MCLOSE + 5:
6363 case MCLOSE + 6:
6364 case MCLOSE + 7:
6365 case MCLOSE + 8:
6366 case MCLOSE + 9:
6367 sprintf(buf + STRLEN(buf), "MCLOSE%d", OP(op) - MCLOSE);
6368 p = NULL;
6369 break;
6370 case BACKREF + 1:
6371 case BACKREF + 2:
6372 case BACKREF + 3:
6373 case BACKREF + 4:
6374 case BACKREF + 5:
6375 case BACKREF + 6:
6376 case BACKREF + 7:
6377 case BACKREF + 8:
6378 case BACKREF + 9:
6379 sprintf(buf + STRLEN(buf), "BACKREF%d", OP(op) - BACKREF);
6380 p = NULL;
6381 break;
6382 case NOPEN:
6383 p = "NOPEN";
6384 break;
6385 case NCLOSE:
6386 p = "NCLOSE";
6387 break;
6388 #ifdef FEAT_SYN_HL
6389 case ZOPEN + 1:
6390 case ZOPEN + 2:
6391 case ZOPEN + 3:
6392 case ZOPEN + 4:
6393 case ZOPEN + 5:
6394 case ZOPEN + 6:
6395 case ZOPEN + 7:
6396 case ZOPEN + 8:
6397 case ZOPEN + 9:
6398 sprintf(buf + STRLEN(buf), "ZOPEN%d", OP(op) - ZOPEN);
6399 p = NULL;
6400 break;
6401 case ZCLOSE + 1:
6402 case ZCLOSE + 2:
6403 case ZCLOSE + 3:
6404 case ZCLOSE + 4:
6405 case ZCLOSE + 5:
6406 case ZCLOSE + 6:
6407 case ZCLOSE + 7:
6408 case ZCLOSE + 8:
6409 case ZCLOSE + 9:
6410 sprintf(buf + STRLEN(buf), "ZCLOSE%d", OP(op) - ZCLOSE);
6411 p = NULL;
6412 break;
6413 case ZREF + 1:
6414 case ZREF + 2:
6415 case ZREF + 3:
6416 case ZREF + 4:
6417 case ZREF + 5:
6418 case ZREF + 6:
6419 case ZREF + 7:
6420 case ZREF + 8:
6421 case ZREF + 9:
6422 sprintf(buf + STRLEN(buf), "ZREF%d", OP(op) - ZREF);
6423 p = NULL;
6424 break;
6425 #endif
6426 case STAR:
6427 p = "STAR";
6428 break;
6429 case PLUS:
6430 p = "PLUS";
6431 break;
6432 case NOMATCH:
6433 p = "NOMATCH";
6434 break;
6435 case MATCH:
6436 p = "MATCH";
6437 break;
6438 case BEHIND:
6439 p = "BEHIND";
6440 break;
6441 case NOBEHIND:
6442 p = "NOBEHIND";
6443 break;
6444 case SUBPAT:
6445 p = "SUBPAT";
6446 break;
6447 case BRACE_LIMITS:
6448 p = "BRACE_LIMITS";
6449 break;
6450 case BRACE_SIMPLE:
6451 p = "BRACE_SIMPLE";
6452 break;
6453 case BRACE_COMPLEX + 0:
6454 case BRACE_COMPLEX + 1:
6455 case BRACE_COMPLEX + 2:
6456 case BRACE_COMPLEX + 3:
6457 case BRACE_COMPLEX + 4:
6458 case BRACE_COMPLEX + 5:
6459 case BRACE_COMPLEX + 6:
6460 case BRACE_COMPLEX + 7:
6461 case BRACE_COMPLEX + 8:
6462 case BRACE_COMPLEX + 9:
6463 sprintf(buf + STRLEN(buf), "BRACE_COMPLEX%d", OP(op) - BRACE_COMPLEX);
6464 p = NULL;
6465 break;
6466 #ifdef FEAT_MBYTE
6467 case MULTIBYTECODE:
6468 p = "MULTIBYTECODE";
6469 break;
6470 #endif
6471 case NEWL:
6472 p = "NEWL";
6473 break;
6474 default:
6475 sprintf(buf + STRLEN(buf), "corrupt %d", OP(op));
6476 p = NULL;
6477 break;
6478 }
6479 if (p != NULL)
6480 (void) strcat(buf, p);
6481 return buf;
6482 }
6483 #endif
6484
6485 #ifdef FEAT_MBYTE
6486 static void mb_decompose __ARGS((int c, int *c1, int *c2, int *c3));
6487
6488 typedef struct
6489 {
6490 int a, b, c;
6491 } decomp_T;
6492
6493
6494 /* 0xfb20 - 0xfb4f */
6495 static decomp_T decomp_table[0xfb4f-0xfb20+1] =
6496 {
6497 {0x5e2,0,0}, /* 0xfb20 alt ayin */
6498 {0x5d0,0,0}, /* 0xfb21 alt alef */
6499 {0x5d3,0,0}, /* 0xfb22 alt dalet */
6500 {0x5d4,0,0}, /* 0xfb23 alt he */
6501 {0x5db,0,0}, /* 0xfb24 alt kaf */
6502 {0x5dc,0,0}, /* 0xfb25 alt lamed */
6503 {0x5dd,0,0}, /* 0xfb26 alt mem-sofit */
6504 {0x5e8,0,0}, /* 0xfb27 alt resh */
6505 {0x5ea,0,0}, /* 0xfb28 alt tav */
6506 {'+', 0, 0}, /* 0xfb29 alt plus */
6507 {0x5e9, 0x5c1, 0}, /* 0xfb2a shin+shin-dot */
6508 {0x5e9, 0x5c2, 0}, /* 0xfb2b shin+sin-dot */
6509 {0x5e9, 0x5c1, 0x5bc}, /* 0xfb2c shin+shin-dot+dagesh */
6510 {0x5e9, 0x5c2, 0x5bc}, /* 0xfb2d shin+sin-dot+dagesh */
6511 {0x5d0, 0x5b7, 0}, /* 0xfb2e alef+patah */
6512 {0x5d0, 0x5b8, 0}, /* 0xfb2f alef+qamats */
6513 {0x5d0, 0x5b4, 0}, /* 0xfb30 alef+hiriq */
6514 {0x5d1, 0x5bc, 0}, /* 0xfb31 bet+dagesh */
6515 {0x5d2, 0x5bc, 0}, /* 0xfb32 gimel+dagesh */
6516 {0x5d3, 0x5bc, 0}, /* 0xfb33 dalet+dagesh */
6517 {0x5d4, 0x5bc, 0}, /* 0xfb34 he+dagesh */
6518 {0x5d5, 0x5bc, 0}, /* 0xfb35 vav+dagesh */
6519 {0x5d6, 0x5bc, 0}, /* 0xfb36 zayin+dagesh */
6520 {0xfb37, 0, 0}, /* 0xfb37 -- UNUSED */
6521 {0x5d8, 0x5bc, 0}, /* 0xfb38 tet+dagesh */
6522 {0x5d9, 0x5bc, 0}, /* 0xfb39 yud+dagesh */
6523 {0x5da, 0x5bc, 0}, /* 0xfb3a kaf sofit+dagesh */
6524 {0x5db, 0x5bc, 0}, /* 0xfb3b kaf+dagesh */
6525 {0x5dc, 0x5bc, 0}, /* 0xfb3c lamed+dagesh */
6526 {0xfb3d, 0, 0}, /* 0xfb3d -- UNUSED */
6527 {0x5de, 0x5bc, 0}, /* 0xfb3e mem+dagesh */
6528 {0xfb3f, 0, 0}, /* 0xfb3f -- UNUSED */
6529 {0x5e0, 0x5bc, 0}, /* 0xfb40 nun+dagesh */
6530 {0x5e1, 0x5bc, 0}, /* 0xfb41 samech+dagesh */
6531 {0xfb42, 0, 0}, /* 0xfb42 -- UNUSED */
6532 {0x5e3, 0x5bc, 0}, /* 0xfb43 pe sofit+dagesh */
6533 {0x5e4, 0x5bc,0}, /* 0xfb44 pe+dagesh */
6534 {0xfb45, 0, 0}, /* 0xfb45 -- UNUSED */
6535 {0x5e6, 0x5bc, 0}, /* 0xfb46 tsadi+dagesh */
6536 {0x5e7, 0x5bc, 0}, /* 0xfb47 qof+dagesh */
6537 {0x5e8, 0x5bc, 0}, /* 0xfb48 resh+dagesh */
6538 {0x5e9, 0x5bc, 0}, /* 0xfb49 shin+dagesh */
6539 {0x5ea, 0x5bc, 0}, /* 0xfb4a tav+dagesh */
6540 {0x5d5, 0x5b9, 0}, /* 0xfb4b vav+holam */
6541 {0x5d1, 0x5bf, 0}, /* 0xfb4c bet+rafe */
6542 {0x5db, 0x5bf, 0}, /* 0xfb4d kaf+rafe */
6543 {0x5e4, 0x5bf, 0}, /* 0xfb4e pe+rafe */
6544 {0x5d0, 0x5dc, 0} /* 0xfb4f alef-lamed */
6545 };
6546
6547 static void
6548 mb_decompose(c, c1, c2, c3)
6549 int c, *c1, *c2, *c3;
6550 {
6551 decomp_T d;
6552
6553 if (c >= 0x4b20 && c <= 0xfb4f)
6554 {
6555 d = decomp_table[c - 0xfb20];
6556 *c1 = d.a;
6557 *c2 = d.b;
6558 *c3 = d.c;
6559 }
6560 else
6561 {
6562 *c1 = c;
6563 *c2 = *c3 = 0;
6564 }
6565 }
6566 #endif
6567
6568 /*
6569 * Compare two strings, ignore case if ireg_ic set.
6570 * Return 0 if strings match, non-zero otherwise.
6571 * Correct the length "*n" when composing characters are ignored.
6572 */
6573 static int
6574 cstrncmp(s1, s2, n)
6575 char_u *s1, *s2;
6576 int *n;
6577 {
6578 int result;
6579
6580 if (!ireg_ic)
6581 result = STRNCMP(s1, s2, *n);
6582 else
6583 result = MB_STRNICMP(s1, s2, *n);
6584
6585 #ifdef FEAT_MBYTE
6586 /* if it failed and it's utf8 and we want to combineignore: */
6587 if (result != 0 && enc_utf8 && ireg_icombine)
6588 {
6589 char_u *str1, *str2;
6590 int c1, c2, c11, c12;
6591 int junk;
6592
6593 /* we have to handle the strcmp ourselves, since it is necessary to
6594 * deal with the composing characters by ignoring them: */
6595 str1 = s1;
6596 str2 = s2;
6597 c1 = c2 = 0;
6598 while ((int)(str1 - s1) < *n)
6599 {
6600 c1 = mb_ptr2char_adv(&str1);
6601 c2 = mb_ptr2char_adv(&str2);
6602
6603 /* decompose the character if necessary, into 'base' characters
6604 * because I don't care about Arabic, I will hard-code the Hebrew
6605 * which I *do* care about! So sue me... */
6606 if (c1 != c2 && (!ireg_ic || utf_fold(c1) != utf_fold(c2)))
6607 {
6608 /* decomposition necessary? */
6609 mb_decompose(c1, &c11, &junk, &junk);
6610 mb_decompose(c2, &c12, &junk, &junk);
6611 c1 = c11;
6612 c2 = c12;
6613 if (c11 != c12 && (!ireg_ic || utf_fold(c11) != utf_fold(c12)))
6614 break;
6615 }
6616 }
6617 result = c2 - c1;
6618 if (result == 0)
6619 *n = (int)(str2 - s2);
6620 }
6621 #endif
6622
6623 return result;
6624 }
6625
6626 /*
6627 * cstrchr: This function is used a lot for simple searches, keep it fast!
6628 */
6629 static char_u *
6630 cstrchr(s, c)
6631 char_u *s;
6632 int c;
6633 {
6634 char_u *p;
6635 int cc;
6636
6637 if (!ireg_ic
6638 #ifdef FEAT_MBYTE
6639 || (!enc_utf8 && mb_char2len(c) > 1)
6640 #endif
6641 )
6642 return vim_strchr(s, c);
6643
6644 /* tolower() and toupper() can be slow, comparing twice should be a lot
6645 * faster (esp. when using MS Visual C++!).
6646 * For UTF-8 need to use folded case. */
6647 #ifdef FEAT_MBYTE
6648 if (enc_utf8 && c > 0x80)
6649 cc = utf_fold(c);
6650 else
6651 #endif
6652 if (MB_ISUPPER(c))
6653 cc = MB_TOLOWER(c);
6654 else if (MB_ISLOWER(c))
6655 cc = MB_TOUPPER(c);
6656 else
6657 return vim_strchr(s, c);
6658
6659 #ifdef FEAT_MBYTE
6660 if (has_mbyte)
6661 {
6662 for (p = s; *p != NUL; p += (*mb_ptr2len)(p))
6663 {
6664 if (enc_utf8 && c > 0x80)
6665 {
6666 if (utf_fold(utf_ptr2char(p)) == cc)
6667 return p;
6668 }
6669 else if (*p == c || *p == cc)
6670 return p;
6671 }
6672 }
6673 else
6674 #endif
6675 /* Faster version for when there are no multi-byte characters. */
6676 for (p = s; *p != NUL; ++p)
6677 if (*p == c || *p == cc)
6678 return p;
6679
6680 return NULL;
6681 }
6682
6683 /***************************************************************
6684 * regsub stuff *
6685 ***************************************************************/
6686
6687 /* This stuff below really confuses cc on an SGI -- webb */
6688 #ifdef __sgi
6689 # undef __ARGS
6690 # define __ARGS(x) ()
6691 #endif
6692
6693 /*
6694 * We should define ftpr as a pointer to a function returning a pointer to
6695 * a function returning a pointer to a function ...
6696 * This is impossible, so we declare a pointer to a function returning a
6697 * pointer to a function returning void. This should work for all compilers.
6698 */
6699 typedef void (*(*fptr_T) __ARGS((int *, int)))();
6700
6701 static fptr_T do_upper __ARGS((int *, int));
6702 static fptr_T do_Upper __ARGS((int *, int));
6703 static fptr_T do_lower __ARGS((int *, int));
6704 static fptr_T do_Lower __ARGS((int *, int));
6705
6706 static int vim_regsub_both __ARGS((char_u *source, char_u *dest, int copy, int magic, int backslash));
6707
6708 static fptr_T
6709 do_upper(d, c)
6710 int *d;
6711 int c;
6712 {
6713 *d = MB_TOUPPER(c);
6714
6715 return (fptr_T)NULL;
6716 }
6717
6718 static fptr_T
6719 do_Upper(d, c)
6720 int *d;
6721 int c;
6722 {
6723 *d = MB_TOUPPER(c);
6724
6725 return (fptr_T)do_Upper;
6726 }
6727
6728 static fptr_T
6729 do_lower(d, c)
6730 int *d;
6731 int c;
6732 {
6733 *d = MB_TOLOWER(c);
6734
6735 return (fptr_T)NULL;
6736 }
6737
6738 static fptr_T
6739 do_Lower(d, c)
6740 int *d;
6741 int c;
6742 {
6743 *d = MB_TOLOWER(c);
6744
6745 return (fptr_T)do_Lower;
6746 }
6747
6748 /*
6749 * regtilde(): Replace tildes in the pattern by the old pattern.
6750 *
6751 * Short explanation of the tilde: It stands for the previous replacement
6752 * pattern. If that previous pattern also contains a ~ we should go back a
6753 * step further... But we insert the previous pattern into the current one
6754 * and remember that.
6755 * This still does not handle the case where "magic" changes. So require the
6756 * user to keep his hands off of "magic".
6757 *
6758 * The tildes are parsed once before the first call to vim_regsub().
6759 */
6760 char_u *
6761 regtilde(source, magic)
6762 char_u *source;
6763 int magic;
6764 {
6765 char_u *newsub = source;
6766 char_u *tmpsub;
6767 char_u *p;
6768 int len;
6769 int prevlen;
6770
6771 for (p = newsub; *p; ++p)
6772 {
6773 if ((*p == '~' && magic) || (*p == '\\' && *(p + 1) == '~' && !magic))
6774 {
6775 if (reg_prev_sub != NULL)
6776 {
6777 /* length = len(newsub) - 1 + len(prev_sub) + 1 */
6778 prevlen = (int)STRLEN(reg_prev_sub);
6779 tmpsub = alloc((unsigned)(STRLEN(newsub) + prevlen));
6780 if (tmpsub != NULL)
6781 {
6782 /* copy prefix */
6783 len = (int)(p - newsub); /* not including ~ */
6784 mch_memmove(tmpsub, newsub, (size_t)len);
6785 /* interpret tilde */
6786 mch_memmove(tmpsub + len, reg_prev_sub, (size_t)prevlen);
6787 /* copy postfix */
6788 if (!magic)
6789 ++p; /* back off \ */
6790 STRCPY(tmpsub + len + prevlen, p + 1);
6791
6792 if (newsub != source) /* already allocated newsub */
6793 vim_free(newsub);
6794 newsub = tmpsub;
6795 p = newsub + len + prevlen;
6796 }
6797 }
6798 else if (magic)
6799 STRMOVE(p, p + 1); /* remove '~' */
6800 else
6801 STRMOVE(p, p + 2); /* remove '\~' */
6802 --p;
6803 }
6804 else
6805 {
6806 if (*p == '\\' && p[1]) /* skip escaped characters */
6807 ++p;
6808 #ifdef FEAT_MBYTE
6809 if (has_mbyte)
6810 p += (*mb_ptr2len)(p) - 1;
6811 #endif
6812 }
6813 }
6814
6815 vim_free(reg_prev_sub);
6816 if (newsub != source) /* newsub was allocated, just keep it */
6817 reg_prev_sub = newsub;
6818 else /* no ~ found, need to save newsub */
6819 reg_prev_sub = vim_strsave(newsub);
6820 return newsub;
6821 }
6822
6823 #ifdef FEAT_EVAL
6824 static int can_f_submatch = FALSE; /* TRUE when submatch() can be used */
6825
6826 /* These pointers are used instead of reg_match and reg_mmatch for
6827 * reg_submatch(). Needed for when the substitution string is an expression
6828 * that contains a call to substitute() and submatch(). */
6829 static regmatch_T *submatch_match;
6830 static regmmatch_T *submatch_mmatch;
6831 static linenr_T submatch_firstlnum;
6832 static linenr_T submatch_maxline;
6833 #endif
6834
6835 #if defined(FEAT_MODIFY_FNAME) || defined(FEAT_EVAL) || defined(PROTO)
6836 /*
6837 * vim_regsub() - perform substitutions after a vim_regexec() or
6838 * vim_regexec_multi() match.
6839 *
6840 * If "copy" is TRUE really copy into "dest".
6841 * If "copy" is FALSE nothing is copied, this is just to find out the length
6842 * of the result.
6843 *
6844 * If "backslash" is TRUE, a backslash will be removed later, need to double
6845 * them to keep them, and insert a backslash before a CR to avoid it being
6846 * replaced with a line break later.
6847 *
6848 * Note: The matched text must not change between the call of
6849 * vim_regexec()/vim_regexec_multi() and vim_regsub()! It would make the back
6850 * references invalid!
6851 *
6852 * Returns the size of the replacement, including terminating NUL.
6853 */
6854 int
6855 vim_regsub(rmp, source, dest, copy, magic, backslash)
6856 regmatch_T *rmp;
6857 char_u *source;
6858 char_u *dest;
6859 int copy;
6860 int magic;
6861 int backslash;
6862 {
6863 reg_match = rmp;
6864 reg_mmatch = NULL;
6865 reg_maxline = 0;
6866 return vim_regsub_both(source, dest, copy, magic, backslash);
6867 }
6868 #endif
6869
6870 int
6871 vim_regsub_multi(rmp, lnum, source, dest, copy, magic, backslash)
6872 regmmatch_T *rmp;
6873 linenr_T lnum;
6874 char_u *source;
6875 char_u *dest;
6876 int copy;
6877 int magic;
6878 int backslash;
6879 {
6880 reg_match = NULL;
6881 reg_mmatch = rmp;
6882 reg_buf = curbuf; /* always works on the current buffer! */
6883 reg_firstlnum = lnum;
6884 reg_maxline = curbuf->b_ml.ml_line_count - lnum;
6885 return vim_regsub_both(source, dest, copy, magic, backslash);
6886 }
6887
6888 static int
6889 vim_regsub_both(source, dest, copy, magic, backslash)
6890 char_u *source;
6891 char_u *dest;
6892 int copy;
6893 int magic;
6894 int backslash;
6895 {
6896 char_u *src;
6897 char_u *dst;
6898 char_u *s;
6899 int c;
6900 int cc;
6901 int no = -1;
6902 fptr_T func = (fptr_T)NULL;
6903 linenr_T clnum = 0; /* init for GCC */
6904 int len = 0; /* init for GCC */
6905 #ifdef FEAT_EVAL
6906 static char_u *eval_result = NULL;
6907 #endif
6908
6909 /* Be paranoid... */
6910 if (source == NULL || dest == NULL)
6911 {
6912 EMSG(_(e_null));
6913 return 0;
6914 }
6915 if (prog_magic_wrong())
6916 return 0;
6917 src = source;
6918 dst = dest;
6919
6920 /*
6921 * When the substitute part starts with "\=" evaluate it as an expression.
6922 */
6923 if (source[0] == '\\' && source[1] == '='
6924 #ifdef FEAT_EVAL
6925 && !can_f_submatch /* can't do this recursively */
6926 #endif
6927 )
6928 {
6929 #ifdef FEAT_EVAL
6930 /* To make sure that the length doesn't change between checking the
6931 * length and copying the string, and to speed up things, the
6932 * resulting string is saved from the call with "copy" == FALSE to the
6933 * call with "copy" == TRUE. */
6934 if (copy)
6935 {
6936 if (eval_result != NULL)
6937 {
6938 STRCPY(dest, eval_result);
6939 dst += STRLEN(eval_result);
6940 vim_free(eval_result);
6941 eval_result = NULL;
6942 }
6943 }
6944 else
6945 {
6946 win_T *save_reg_win;
6947 int save_ireg_ic;
6948
6949 vim_free(eval_result);
6950
6951 /* The expression may contain substitute(), which calls us
6952 * recursively. Make sure submatch() gets the text from the first
6953 * level. Don't need to save "reg_buf", because
6954 * vim_regexec_multi() can't be called recursively. */
6955 submatch_match = reg_match;
6956 submatch_mmatch = reg_mmatch;
6957 submatch_firstlnum = reg_firstlnum;
6958 submatch_maxline = reg_maxline;
6959 save_reg_win = reg_win;
6960 save_ireg_ic = ireg_ic;
6961 can_f_submatch = TRUE;
6962
6963 eval_result = eval_to_string(source + 2, NULL, TRUE);
6964 if (eval_result != NULL)
6965 {
6966 int had_backslash = FALSE;
6967
6968 for (s = eval_result; *s != NUL; mb_ptr_adv(s))
6969 {
6970 /* Change NL to CR, so that it becomes a line break.
6971 * Skip over a backslashed character. */
6972 if (*s == NL)
6973 *s = CAR;
6974 else if (*s == '\\' && s[1] != NUL)
6975 {
6976 ++s;
6977 /* Change NL to CR here too, so that this works:
6978 * :s/abc\\\ndef/\="aaa\\\nbbb"/ on text:
6979 * abc\
6980 * def
6981 */
6982 if (*s == NL)
6983 *s = CAR;
6984 had_backslash = TRUE;
6985 }
6986 }
6987 if (had_backslash && backslash)
6988 {
6989 /* Backslashes will be consumed, need to double them. */
6990 s = vim_strsave_escaped(eval_result, (char_u *)"\\");
6991 if (s != NULL)
6992 {
6993 vim_free(eval_result);
6994 eval_result = s;
6995 }
6996 }
6997
6998 dst += STRLEN(eval_result);
6999 }
7000
7001 reg_match = submatch_match;
7002 reg_mmatch = submatch_mmatch;
7003 reg_firstlnum = submatch_firstlnum;
7004 reg_maxline = submatch_maxline;
7005 reg_win = save_reg_win;
7006 ireg_ic = save_ireg_ic;
7007 can_f_submatch = FALSE;
7008 }
7009 #endif
7010 }
7011 else
7012 while ((c = *src++) != NUL)
7013 {
7014 if (c == '&' && magic)
7015 no = 0;
7016 else if (c == '\\' && *src != NUL)
7017 {
7018 if (*src == '&' && !magic)
7019 {
7020 ++src;
7021 no = 0;
7022 }
7023 else if ('0' <= *src && *src <= '9')
7024 {
7025 no = *src++ - '0';
7026 }
7027 else if (vim_strchr((char_u *)"uUlLeE", *src))
7028 {
7029 switch (*src++)
7030 {
7031 case 'u': func = (fptr_T)do_upper;
7032 continue;
7033 case 'U': func = (fptr_T)do_Upper;
7034 continue;
7035 case 'l': func = (fptr_T)do_lower;
7036 continue;
7037 case 'L': func = (fptr_T)do_Lower;
7038 continue;
7039 case 'e':
7040 case 'E': func = (fptr_T)NULL;
7041 continue;
7042 }
7043 }
7044 }
7045 if (no < 0) /* Ordinary character. */
7046 {
7047 if (c == K_SPECIAL && src[0] != NUL && src[1] != NUL)
7048 {
7049 /* Copy a special key as-is. */
7050 if (copy)
7051 {
7052 *dst++ = c;
7053 *dst++ = *src++;
7054 *dst++ = *src++;
7055 }
7056 else
7057 {
7058 dst += 3;
7059 src += 2;
7060 }
7061 continue;
7062 }
7063
7064 if (c == '\\' && *src != NUL)
7065 {
7066 /* Check for abbreviations -- webb */
7067 switch (*src)
7068 {
7069 case 'r': c = CAR; ++src; break;
7070 case 'n': c = NL; ++src; break;
7071 case 't': c = TAB; ++src; break;
7072 /* Oh no! \e already has meaning in subst pat :-( */
7073 /* case 'e': c = ESC; ++src; break; */
7074 case 'b': c = Ctrl_H; ++src; break;
7075
7076 /* If "backslash" is TRUE the backslash will be removed
7077 * later. Used to insert a literal CR. */
7078 default: if (backslash)
7079 {
7080 if (copy)
7081 *dst = '\\';
7082 ++dst;
7083 }
7084 c = *src++;
7085 }
7086 }
7087 #ifdef FEAT_MBYTE
7088 else if (has_mbyte)
7089 c = mb_ptr2char(src - 1);
7090 #endif
7091
7092 /* Write to buffer, if copy is set. */
7093 if (func == (fptr_T)NULL) /* just copy */
7094 cc = c;
7095 else
7096 /* Turbo C complains without the typecast */
7097 func = (fptr_T)(func(&cc, c));
7098
7099 #ifdef FEAT_MBYTE
7100 if (has_mbyte)
7101 {
7102 src += mb_ptr2len(src - 1) - 1;
7103 if (copy)
7104 mb_char2bytes(cc, dst);
7105 dst += mb_char2len(cc) - 1;
7106 }
7107 else
7108 #endif
7109 if (copy)
7110 *dst = cc;
7111 dst++;
7112 }
7113 else
7114 {
7115 if (REG_MULTI)
7116 {
7117 clnum = reg_mmatch->startpos[no].lnum;
7118 if (clnum < 0 || reg_mmatch->endpos[no].lnum < 0)
7119 s = NULL;
7120 else
7121 {
7122 s = reg_getline(clnum) + reg_mmatch->startpos[no].col;
7123 if (reg_mmatch->endpos[no].lnum == clnum)
7124 len = reg_mmatch->endpos[no].col
7125 - reg_mmatch->startpos[no].col;
7126 else
7127 len = (int)STRLEN(s);
7128 }
7129 }
7130 else
7131 {
7132 s = reg_match->startp[no];
7133 if (reg_match->endp[no] == NULL)
7134 s = NULL;
7135 else
7136 len = (int)(reg_match->endp[no] - s);
7137 }
7138 if (s != NULL)
7139 {
7140 for (;;)
7141 {
7142 if (len == 0)
7143 {
7144 if (REG_MULTI)
7145 {
7146 if (reg_mmatch->endpos[no].lnum == clnum)
7147 break;
7148 if (copy)
7149 *dst = CAR;
7150 ++dst;
7151 s = reg_getline(++clnum);
7152 if (reg_mmatch->endpos[no].lnum == clnum)
7153 len = reg_mmatch->endpos[no].col;
7154 else
7155 len = (int)STRLEN(s);
7156 }
7157 else
7158 break;
7159 }
7160 else if (*s == NUL) /* we hit NUL. */
7161 {
7162 if (copy)
7163 EMSG(_(e_re_damg));
7164 goto exit;
7165 }
7166 else
7167 {
7168 if (backslash && (*s == CAR || *s == '\\'))
7169 {
7170 /*
7171 * Insert a backslash in front of a CR, otherwise
7172 * it will be replaced by a line break.
7173 * Number of backslashes will be halved later,
7174 * double them here.
7175 */
7176 if (copy)
7177 {
7178 dst[0] = '\\';
7179 dst[1] = *s;
7180 }
7181 dst += 2;
7182 }
7183 else
7184 {
7185 #ifdef FEAT_MBYTE
7186 if (has_mbyte)
7187 c = mb_ptr2char(s);
7188 else
7189 #endif
7190 c = *s;
7191
7192 if (func == (fptr_T)NULL) /* just copy */
7193 cc = c;
7194 else
7195 /* Turbo C complains without the typecast */
7196 func = (fptr_T)(func(&cc, c));
7197
7198 #ifdef FEAT_MBYTE
7199 if (has_mbyte)
7200 {
7201 int l;
7202
7203 /* Copy composing characters separately, one
7204 * at a time. */
7205 if (enc_utf8)
7206 l = utf_ptr2len(s) - 1;
7207 else
7208 l = mb_ptr2len(s) - 1;
7209
7210 s += l;
7211 len -= l;
7212 if (copy)
7213 mb_char2bytes(cc, dst);
7214 dst += mb_char2len(cc) - 1;
7215 }
7216 else
7217 #endif
7218 if (copy)
7219 *dst = cc;
7220 dst++;
7221 }
7222
7223 ++s;
7224 --len;
7225 }
7226 }
7227 }
7228 no = -1;
7229 }
7230 }
7231 if (copy)
7232 *dst = NUL;
7233
7234 exit:
7235 return (int)((dst - dest) + 1);
7236 }
7237
7238 #ifdef FEAT_EVAL
7239 static char_u *reg_getline_submatch __ARGS((linenr_T lnum));
7240
7241 /*
7242 * Call reg_getline() with the line numbers from the submatch. If a
7243 * substitute() was used the reg_maxline and other values have been
7244 * overwritten.
7245 */
7246 static char_u *
7247 reg_getline_submatch(lnum)
7248 linenr_T lnum;
7249 {
7250 char_u *s;
7251 linenr_T save_first = reg_firstlnum;
7252 linenr_T save_max = reg_maxline;
7253
7254 reg_firstlnum = submatch_firstlnum;
7255 reg_maxline = submatch_maxline;
7256
7257 s = reg_getline(lnum);
7258
7259 reg_firstlnum = save_first;
7260 reg_maxline = save_max;
7261 return s;
7262 }
7263
7264 /*
7265 * Used for the submatch() function: get the string from the n'th submatch in
7266 * allocated memory.
7267 * Returns NULL when not in a ":s" command and for a non-existing submatch.
7268 */
7269 char_u *
7270 reg_submatch(no)
7271 int no;
7272 {
7273 char_u *retval = NULL;
7274 char_u *s;
7275 int len;
7276 int round;
7277 linenr_T lnum;
7278
7279 if (!can_f_submatch || no < 0)
7280 return NULL;
7281
7282 if (submatch_match == NULL)
7283 {
7284 /*
7285 * First round: compute the length and allocate memory.
7286 * Second round: copy the text.
7287 */
7288 for (round = 1; round <= 2; ++round)
7289 {
7290 lnum = submatch_mmatch->startpos[no].lnum;
7291 if (lnum < 0 || submatch_mmatch->endpos[no].lnum < 0)
7292 return NULL;
7293
7294 s = reg_getline_submatch(lnum) + submatch_mmatch->startpos[no].col;
7295 if (s == NULL) /* anti-crash check, cannot happen? */
7296 break;
7297 if (submatch_mmatch->endpos[no].lnum == lnum)
7298 {
7299 /* Within one line: take form start to end col. */
7300 len = submatch_mmatch->endpos[no].col
7301 - submatch_mmatch->startpos[no].col;
7302 if (round == 2)
7303 vim_strncpy(retval, s, len);
7304 ++len;
7305 }
7306 else
7307 {
7308 /* Multiple lines: take start line from start col, middle
7309 * lines completely and end line up to end col. */
7310 len = (int)STRLEN(s);
7311 if (round == 2)
7312 {
7313 STRCPY(retval, s);
7314 retval[len] = '\n';
7315 }
7316 ++len;
7317 ++lnum;
7318 while (lnum < submatch_mmatch->endpos[no].lnum)
7319 {
7320 s = reg_getline_submatch(lnum++);
7321 if (round == 2)
7322 STRCPY(retval + len, s);
7323 len += (int)STRLEN(s);
7324 if (round == 2)
7325 retval[len] = '\n';
7326 ++len;
7327 }
7328 if (round == 2)
7329 STRNCPY(retval + len, reg_getline_submatch(lnum),
7330 submatch_mmatch->endpos[no].col);
7331 len += submatch_mmatch->endpos[no].col;
7332 if (round == 2)
7333 retval[len] = NUL;
7334 ++len;
7335 }
7336
7337 if (retval == NULL)
7338 {
7339 retval = lalloc((long_u)len, TRUE);
7340 if (retval == NULL)
7341 return NULL;
7342 }
7343 }
7344 }
7345 else
7346 {
7347 s = submatch_match->startp[no];
7348 if (s == NULL || submatch_match->endp[no] == NULL)
7349 retval = NULL;
7350 else
7351 retval = vim_strnsave(s, (int)(submatch_match->endp[no] - s));
7352 }
7353
7354 return retval;
7355 }
7356 #endif
Vim with various extensions